# OpenSoftware > OpenSoftware builds private, local-first AI products for everyday work. ## Product context Clovy is OpenSoftware's private AI workspace for macOS. It combines chat, voice dictation into any app, automated meeting notes, and a local agent in one desktop product. Clovy is local-first and open source, with privacy-preserving inference paths through Venice AI and a local agent built on Hermes Agent. Primary URLs: - Website: https://www.opensoftware.co/ - Clovy product page: https://www.opensoftware.co/clovy - Clovy changelog: https://www.opensoftware.co/clovy/changelog - Clovy changelog Markdown: https://www.opensoftware.co/clovy/changelog/content.md - Download: https://www.opensoftware.co/download/mac - Community: https://t.me/+B4Z8KUqEsRc4ZGVh - Verify backend attestation: https://www.opensoftware.co/verify - Newsroom: https://www.opensoftware.co/news - RSS: https://www.opensoftware.co/news/feed.xml - GitHub: https://github.com/open-software-network/os-clovy - Releases: https://github.com/open-software-network/os-june-releases/releases - Venice AI: https://venice.ai/ - Hermes Agent: https://hermes-agent.nousresearch.com/ ## Comparisons ### Clovy vs Granola URL: https://www.opensoftware.co/clovy/vs/granola (updated 2026-07-02) The short answer: Choose Clovy if you want meeting notes stored on your Mac instead of a company's cloud, no model training on your content, and dictation, chat, and a local agent in the same app for the same money. Granola makes sense if your team needs shared notes on Windows and mobile. Both capture meetings without a bot; only one keeps them on your device. Clovy and Granola both take meeting notes without sending a bot into your call. The comparison starts after the meeting ends, and there it stops being close. Granola uploads your transcripts and notes to its US cloud servers and trains its own models on de-identified data by default. Clovy stores recordings, transcripts, and notes on your Mac, routes model calls to zero-retention models by default, and never trains on your data. And meeting notes are one part of Clovy's private workspace, next to dictation, chat, and a local agent. **Same capture method, different storage model** Both apps capture meetings from your computer's audio, so neither sends a bot into your call. On capture, they are even. On storage, they are not. Granola processes and stores your transcripts and notes on US cloud servers; it discards the raw audio, but everything said in your meetings still lives on its infrastructure, under its retention policy. Clovy stores recordings, transcripts, and notes on your Mac. Audio goes out for note transcription through zero-retention model routing by default: nothing stored, nothing trained on, and OpenSoftware keeps only account, login, and billing records. **Read the training defaults before you pick** Granola's own privacy policy says it trains its models on de-identified data, on by default for individuals, with an opt-out in settings. That means trusting both the toggle and the de-identification method with everything said in your meetings. It does bar third parties like OpenAI and Anthropic from training on your data, but Granola itself is the exception. Clovy has no exception. OpenSoftware never trains on your data, the default model routing is zero retention, and the backend is open source and TEE-attested, so the claim is checkable rather than a policy promise. **A notetaker vs a workspace** Granola is a focused meeting tool, and focus has benefits. Clovy's bet is that meeting notes belong next to the rest of your AI work: dictation that types into any app, chat for thinking things through, and an agent that runs on your Mac and can use your meetings and notes as context. One private workspace instead of one more subscription per feature. Q: Do Clovy and Granola both work without a meeting bot? A: Yes. Both capture meetings from your computer's audio, so no bot joins the call in either app. The difference is what happens after: Granola stores transcripts and notes on US cloud servers, while Clovy stores recordings, transcripts, and notes on your Mac. Q: Is Granola private? A: Less private than it first appears. Granola discards raw audio, but your transcripts and notes are stored on its US cloud servers, and its privacy policy says it trains its own models on de-identified data by default unless you opt out. Clovy stores your meetings on your Mac, routes model calls to zero-retention models by default, and never trains on your data. Q: Which is cheaper, Clovy or Granola? A: Both have a free plan. Granola's free plan limits you to a rolling 30 day window of notes; Clovy's free plan keeps your full history because notes live on your Mac. Paid is $14 per user per month for Granola Business and $20 per month for Clovy Pro, which also covers dictation and the agent. Q: Can Clovy import my Granola notes? A: There is no one-click import today. Granola lets you export notes, and Clovy's agent can help you organize exported files on your Mac. Your Clovy notes are stored on your device, so leaving Clovy never requires an export in the first place. --- ### Clovy vs Superwhisper URL: https://www.opensoftware.co/clovy/vs/superwhisper (updated 2026-07-02) The short answer: Choose Clovy if you want dictation that is private by default on every plan, including free: zero-retention model routing with nothing to configure, history on your Mac, open-source code you can verify, plus meeting notes and a local agent in the same app. Superwhisper fits one specific requirement Clovy does not meet: with a paid license on Apple Silicon, it can run fully on device. Clovy and Superwhisper are both built for people who care where their voice data goes. They just answer differently. Superwhisper's answer is configuration: pick the right models from its catalog, on the right hardware, with a Pro license, and dictation can run entirely on device. Clovy's answer is defaults: every dictation on every plan routes to zero-retention models with nothing to set up, nothing stored, nothing trained on. Defaults are the difference that matters in practice, because most people never change them. Clovy's private path is the one you are on the moment it is installed. And dictation is one part of Clovy's workspace, next to meeting notes with no bot, chat, and a local agent. **Private in expert hands vs private out of the box** The concession first: with a Pro license on Apple Silicon, Superwhisper can run a local voice model and a local language model, and that dictation never touches a server. Clovy does not offer an on-device mode, and we will not pretend otherwise. Now the fine print that page rarely leads with. That mode is one configuration among many: the free tier and the highest quality models route to the cloud, offline models need Apple Silicon to run well, and what actually leaves your Mac depends on which of dozens of models you picked. Superwhisper can be private in expert hands. A default is not what it has. Clovy's bet is that defaults protect more people than options do. Every dictation on every plan routes to zero-retention models: nothing stored, nothing trained on, no setup, no model menu to study first. Your history stays on your Mac, OpenSoftware keeps only account, login, and billing records, and the routing backend is open source and TEE-attested, so the claim is verifiable instead of configurable. **A dictation tool vs a private workspace** Superwhisper is a dictation tool, and only that: it does not detect your meetings, take structured notes, answer questions, or run tasks. Clovy's dictation is one surface of a private workspace. It detects your meetings and writes notes without a bot, answers questions in chat, and runs an agent on your Mac that can act on all of it. One app, one subscription, one privacy standard across everything you say to it. Q: Is Superwhisper more private than Clovy? A: Only in one specific setup: Superwhisper Pro on Apple Silicon, manually configured with local models, keeps audio fully on device, which Clovy does not match. In every other configuration the comparison flips. Superwhisper's free tier and best models route to the cloud and privacy depends on your choices, while Clovy routes every plan to zero-retention models by default, keeps your history on your Mac, and is open source so you can verify it. Q: Does Clovy have on-device transcription models? A: No. Clovy sends dictation audio out for transcription and cleanup through zero-retention model routing by default, meaning nothing is stored and nothing is trained on. Your dictation history stays on your Mac. Q: Which is cheaper? A: Superwhisper charges less for dictation alone: $8.49 per month or $249.99 lifetime. Clovy Pro is $20 per month for the whole workspace it would take several tools to match: dictation, meeting notes, chat, and a local agent, all under one privacy standard. Both have free tiers; Clovy's free tier carries the same privacy standard as Pro. Q: Do both apps handle meetings? A: Both can record and summarize a meeting without a bot joining the call. Clovy detects supported meetings and offers to write structured notes automatically, and stores recordings, transcripts, and notes on your Mac. Superwhisper's meeting mode is a manual recording mode with a summary. --- ### Clovy vs Otter.ai URL: https://www.opensoftware.co/clovy/vs/otter (updated 2026-07-02) The short answer: Choose Clovy if you do not want a bot in your meetings or your conversations stored and trained on by a transcription company: Clovy captures from your Mac's audio, keeps transcripts on your device, never trains on your data, and includes dictation and a local agent. Otter fits if you need a live transcript with named speakers during the call, or non-Mac platforms, and accept the cloud that comes with it. Everything Otter transcribes goes to its US cloud servers, and its own privacy pages say those conversations train its models, automatically. That is the deal behind the familiar bot in the participant list: your meetings become its data. Clovy is built on the opposite premise: your meetings are not a dataset. Clovy takes notes from your Mac's audio with no bot in the call, stores recordings, transcripts, and notes on your device, and never trains on your data. It also goes beyond meetings, with dictation and a local agent in the same private workspace. **The bot is the business model** Otter works primarily through a notetaker bot that joins your calls via calendar sync, and it is famous for showing up uninvited when default settings are left on. A no-bot desktop capture mode exists, but the bot is the product's center of gravity, and either way your conversations end up on Otter's servers. Clovy never has a bot. It detects supported meetings on your Mac, asks if you want notes, and captures from your device's audio. Nothing joins the call, and nothing appears in the participant list. **Where a decade of conversations lives** Everything Otter transcribes is processed and stored on its US cloud servers, and its privacy pages state that it uses de-identified recordings and transcripts to train its models, automatically. You are trusting both the de-identification method and the retention policy. Clovy stores recordings, transcripts, and notes on your Mac. Audio goes out for note transcription through zero-retention model routing by default, so nothing is stored and nothing is trained on. OpenSoftware keeps only account, login, and billing records, and the backend is open source and TEE-attested so you can verify that instead of trusting it. **The honest tradeoff** One concession: if you need a transcript scrolling in real time with named speakers, Otter does that and Clovy does not. Clovy writes its notes when the meeting ends, and it segments the transcript by turns rather than naming speakers. Weigh that against everything else on this page: where a decade of your conversations lives, whether they train someone's models, and what you get around the meeting. Clovy gives you dictation into any app, private chat, and an agent on your Mac that can use your notes as context, all without your meetings leaving your device. Q: Does Clovy join meetings with a bot like Otter does? A: No. Clovy captures audio from your Mac, so no bot ever joins your call. Otter primarily uses a notetaker bot that joins meetings through your calendar, though it also offers desktop capture on Mac and Windows. Q: Does Otter train AI models on my meetings? A: Otter's own privacy pages say it uses de-identified recordings and transcripts to train its models, and that this happens automatically. Clovy never trains on your data, and its default model routing is zero retention. Q: Does Clovy have live transcription like Otter? A: No. Clovy records the meeting on your Mac and writes the transcript and structured notes when it ends, stored on your device. Otter offers a live scrolling transcript with named speakers during the call; the price of that pipeline is that your conversation streams through Otter's cloud. Q: Is Clovy cheaper than Otter? A: They are close. Otter Pro is $16.99 per month, or $8.33 per month billed annually, and covers transcription. Clovy Pro is $20 per month and covers meeting notes plus dictation, chat, and the agent. Both have free plans; Clovy's free plan keeps everything on your Mac. --- ### Clovy vs Wispr Flow URL: https://www.opensoftware.co/clovy/vs/wispr-flow (updated 2026-07-02) The short answer: Choose Clovy if you want dictation that is private without configuration: Clovy routes to zero-retention models by default, keeps your history on your Mac, never trains on your data, and includes meeting notes and a local agent in the same app. Wispr Flow fits if you need dictation on Windows or your phone and accept that your voice data is stored in the cloud and used for training until you change the settings. Before you pick a dictation app, read its data page. Wispr Flow's says transcription always happens in the cloud, and by default your audio and transcripts are stored on its servers and may be used to train its models. The private configuration exists, but you have to find the toggles, and most people never do. Clovy does not ask you to. Dictation routes to zero-retention models by default on every plan: nothing stored, nothing trained on, history on your Mac. And it comes inside a private workspace with meeting notes, chat, and a local agent, not as one more single-purpose subscription. **Defaults are the real privacy policy** Wispr Flow's own data controls page spells it out: transcription always happens in the cloud, privacy mode is off by default, cloud sync is on, and your audio and transcripts are stored on US servers and may be used to train its models until you find the toggles. A locked-down configuration exists, and it is the default only for enterprise customers. Everyone else starts on the data-collecting path. Clovy does not make you configure privacy. Dictation routes to zero-retention models by default on every plan: nothing stored, nothing trained on. Your history stays on your Mac, OpenSoftware keeps only account, login, and billing records, and the routing backend is open source and TEE-attested, so you can verify all of this instead of trusting a settings page. **What your $15 or $20 buys** Wispr Flow Pro is $15 per month for one thing: dictation. Clovy Pro is $20 per month for the whole workspace: dictation into any app, meeting notes with no bot, private chat, and an agent that runs on your Mac. If dictation is one of several things you want AI for, one private subscription replaces a stack of cloud ones. Q: Is Wispr Flow private? A: Not by default. For standard accounts, Wispr Flow's privacy mode is off and cloud sync is on, so your audio and transcripts are stored on its servers and may be used for training until you change the settings. Zero retention is default only for enterprise accounts. Clovy's defaults are the private path on every plan: zero-retention routing, no training, history on your Mac. Q: Does Clovy's dictation run on device like some Mac tools? A: No. Clovy sends dictation audio out for transcription and cleanup, routed to zero-retention models by default, so nothing is stored and nothing is trained on. Your dictation history stays on your Mac. If you require transcription that never leaves the device, see our Superwhisper comparison. Q: Which has the better free plan for dictation? A: Wispr Flow's free plan caps you at 2,000 words per week on desktop. Clovy's free Hobby plan is limited by overall usage rather than a weekly word count, and it also includes meeting notes and light agent usage, with the same privacy standard as Pro. Q: Does Wispr Flow do meeting notes or agent tasks? A: No. Wispr Flow is focused on voice input, with extras like a scratchpad and text transforms. Clovy includes automated meeting notes with no bot and a local agent alongside dictation. --- ### Clovy vs ChatGPT desktop URL: https://www.opensoftware.co/clovy/vs/chatgpt-desktop (updated 2026-07-02) The short answer: Choose Clovy if you want a desktop AI workspace where the agent runs on your Mac, your conversations stay on your disk, model calls route to zero-retention models by default, and the code is open source so you can verify all of it. ChatGPT fits if you want OpenAI's frontier models and ecosystem, and accept that your chats live on OpenAI's servers and train its models by default on consumer plans. A desktop AI app sees more of your work than a chatbot tab ever did: your files, your meetings, your voice, your screen. That is exactly why where it sends everything matters more on the desktop than anywhere else. The ChatGPT Mac app is a window into OpenAI's cloud. Your chats are stored there and train its models by default, its agent runs on OpenAI's virtual machines, and app content you share is sent to its servers. Clovy is built the other way around: the agent runs on your Mac, your data stays on your disk, model calls route to zero-retention models by default, and the whole thing is open source. **Where your words go** Every ChatGPT conversation lives on OpenAI's servers, and on consumer plans it is used to improve OpenAI's models unless you find the opt-out. Temporary chats still sit on their servers for up to 30 days. The risk of that architecture stopped being hypothetical in 2025, when a court order in the New York Times case forced OpenAI to preserve consumer chats, including ones users had deleted, and in January 2026 a judge affirmed that 20 million conversations be handed to plaintiffs in de-identified form. None of that is carelessness by OpenAI. It is simply what can happen to data that lives on someone else's servers. Clovy's answer is to hold almost nothing. Chats, notes, transcripts, and agent sessions are stored on your Mac. Model calls route to zero-retention models by default, so nothing is stored and nothing is trained on. OpenSoftware keeps only account, login, and billing records. **Two very different agents** ChatGPT's agent runs entirely on an OpenAI-hosted virtual machine in the cloud. To act on your work, your work has to go to it: files uploaded, context shared, results stored in your ChatGPT history on OpenAI's servers. Clovy's agent is built on the open-source Hermes framework and runs on your Mac, sandboxed by default, with approvals for anything risky. Your files, memory, and sessions never leave your disk; only the model calls do, through zero-retention routing. The agent can make mistakes, and we say so. The guardrails are the point. **The desktop details** Some practical differences add up on a Mac. ChatGPT's desktop app requires Apple Silicon; Clovy also runs on Intel Macs. OpenAI retired voice mode from its Mac app in January 2026; Clovy's dictation types polished text into any app with a keypress. ChatGPT's record mode is reserved for paid plans and stores results on OpenAI's servers; Clovy's meeting notes are part of the free plan, with the recording, transcript, and notes stored on your Mac. Q: Is Clovy a ChatGPT replacement? A: For everyday desktop AI, yes: chat, research, dictation, meeting notes, and agent tasks in one private workspace, with strong zero-retention models by default and a catalog to pick from, each labeled with its privacy tier. The exception is work that depends on one specific OpenAI model; if that is you, you already know it. Q: Does ChatGPT train on my conversations? A: On consumer plans, yes by default: OpenAI uses your chats to improve its models unless you turn that off in data controls. Business and enterprise plans are excluded by default. Clovy never trains on your data on any plan, and its default model routing is zero retention. Q: Does ChatGPT's Mac app have meeting notes? A: Yes. Record mode captures meetings from your Mac and produces a transcript and summary, on paid plans, with the results stored in your ChatGPT history on OpenAI's servers. Clovy's meeting notes are included free and stored on your Mac. Q: What does Clovy use instead of OpenAI's models? A: Clovy defaults to private Venice models with zero data retention and lets you choose from a catalog of models. Every model in the picker shows a privacy tier: E2EE, Private, or Anonymous, so you know exactly what a switch means. ## Use-case guides ### AI meeting notes without a bot URL: https://www.opensoftware.co/clovy/ai-meeting-notes-without-a-bot (updated 2026-07-02) Most AI notetakers join your meeting as a participant. A bot shows up in the call, everyone sees it, and the recording lives on someone else's servers. Clovy works differently. It listens through your Mac's audio, so there is nothing to invite and nothing extra on the call. When Clovy detects a supported meeting, it offers to take notes. You approve it, run the meeting, and get a transcript and structured notes when you are done. **How Clovy captures meetings without a bot** Clovy records from your Mac directly: your microphone, and system audio when you want the other side of the call captured. Because the capture happens on your device, no bot joins the meeting and nothing appears in the participant list. That also means Clovy works wherever you take the call from your Mac. There is no per-app integration to configure and no calendar bot to manage. **Where your meeting data lives** Recordings, transcripts, and notes are stored on your Mac. Audio is sent out for note transcription, and Clovy routes those calls to zero-retention models by default: nothing stored, nothing trained on. OpenSoftware keeps only account, login, and billing records. You do not have to take that on faith. Clovy is open source under the MIT license, and the backend that routes model calls runs in a TEE with cryptographic attestation you can check yourself. **What you get after the meeting** Clovy writes structured notes from the transcript, so you end the call with something you can act on rather than a wall of text. Notes live alongside the transcript in the app. Because Clovy also includes an agent, your meetings become context you can use. Ask follow-up questions, pull decisions out of last week's call, or have Clovy draft the recap message. All of it stays in your private workspace. **Consent still matters** No bot in the call does not mean no responsibility. Depending on where you and the other participants are, you may need consent to record. Clovy leaves that in your hands, and we publish plain guidance in our recording consent page. Telling people you are taking notes is usually the right move anyway. Q: Does Clovy send a bot to join my meetings? A: No. Clovy captures audio from your Mac directly, so nothing joins the call and nothing appears in the participant list. There is no calendar bot to configure or uninvite. Q: Can other people in the meeting see Clovy? A: No. Clovy runs on your Mac and does not appear in the meeting. Participants only know Clovy is taking notes if you tell them, which we recommend where consent rules apply. Q: Where are my meeting recordings and transcripts stored? A: Recordings, transcripts, and notes are stored on your Mac. Audio is sent out for note transcription through zero-retention model routing by default, so nothing is stored on model servers and nothing is trained on. OpenSoftware stores only account, login, and billing records. Q: Does Clovy train AI models on my meetings? A: No. OpenSoftware never trains on your data, and Clovy routes to zero-retention models by default, so the default model providers store nothing and train on nothing. Anonymized third-party models are opt-in and clearly labeled. Q: Is Clovy's meeting notes feature free? A: Yes. Clovy's free Hobby plan includes meeting notes, dictation, and light agent usage, with the same privacy standard as Pro. Pro is $20 per month for higher usage. --- ### Private dictation for Mac URL: https://www.opensoftware.co/clovy/private-dictation-for-mac (updated 2026-07-02) Dictation is the fastest way to write, and the most personal data you can hand an app. Half-formed thoughts, names, numbers, the message you almost sent. Where that audio goes should not be a mystery. Clovy's dictation is built for that reality. Hold the fn key, talk, let go, and polished writing lands in whatever app you are using. Your audio routes to zero-retention models by default, and your dictation history stays on your Mac. **How it works** Clovy's dictation is push-to-talk. Hold the fn key, speak, and release. Clovy turns what you said into clean, punctuated writing and types it where your cursor is. For longer stretches there is a hands-free toggle, and recordings can run long without falling over. It works in any app on your Mac: your email client, your editor, a chat box in the browser, a commit message in the terminal. There is nothing to integrate per app. **Writing that sounds like you** Raw transcription is not the goal. Clovy cleans up filler, fixes punctuation, and applies the style you pick: standard, casual lowercase, or formal. You talk like a person and Clovy writes it like you would have typed it. **What happens to your audio** We are precise about this because most dictation apps are not. When you dictate, your audio leaves your Mac for transcription and cleanup. Clovy routes those calls to zero-retention models by default: nothing stored, nothing trained on. Your dictation history is stored on your Mac, not on our servers. OpenSoftware keeps only account, login, and billing records. The pipeline is verifiable, not just promised. Clovy is open source, and the backend that routes model calls runs in a TEE with cryptographic attestation. You can check what runs on our servers yourself. **More than a dictation app** Clovy is a private AI workspace, so dictation comes with company: automated meeting notes, chat, and a local agent that can act on what you say. If dictation is how you discover Clovy, the rest is already installed. Q: Is Clovy's dictation processed on my Mac? A: The app and your dictation history live on your Mac, and audio is sent out for transcription and cleanup. Clovy routes those calls to zero-retention models by default, meaning nothing is stored and nothing is trained on. If you need transcription that never leaves the device, see our Superwhisper comparison for the honest tradeoff. Q: Which apps does Clovy's dictation work in? A: Any app on your Mac. Clovy types the polished text wherever your cursor is: email, documents, chat apps, code editors, browsers. There is no per-app setup. Q: What is the dictation shortcut? A: Push-to-talk is the fn key by default: hold it, talk, and let go. There is also a hands-free toggle for longer dictation. You can change the shortcut in settings. Q: Does Clovy's dictation cost money? A: Clovy is free to start. The Hobby plan includes dictation with the same privacy standard as Pro. Pro is $20 per month for higher usage across dictation, meeting notes, and the agent. Q: Does Clovy train on my dictation? A: No. OpenSoftware never trains on your data, and the default zero-retention model routing means the providers Clovy uses by default store nothing and train on nothing. Anonymized third-party models are opt-in. --- ### A local AI agent for your Mac URL: https://www.opensoftware.co/clovy/local-ai-agent (updated 2026-07-02) Most AI agents run in the cloud. You hand a remote service your files, your logins, and a window into your work, then hope its security holds. A cloud agent with that reach is a remote company's window into your work, not just your chat box. Clovy takes the other path. The agent runs on your Mac, where your files already are. It is built on Nous Research's open-source Hermes framework, sandboxed by default, and anything risky waits for your yes. **What local means here, precisely** Honesty first: local does not mean nothing ever leaves your Mac. The agent process runs on your device. Your files, sessions, memory, and agent state stay on your disk. When the agent thinks, the model call leaves your Mac, and Clovy routes it to zero-retention models by default: nothing stored, nothing trained on. That split matters. A cloud agent uploads your working context to someone else's computer to act on it. Clovy's agent acts on your computer and sends out only the model calls, with identifying metadata stripped when third-party models are involved. **Sandboxed by default** Fresh agent sessions run in a sandbox: writes are blocked outside the agent's workspace, and a denylist keeps it away from secrets like SSH keys and credentials. Risky actions ask for your approval before they happen. You can grant a session more freedom when a task needs it, and that choice is per session, not a global switch. The agent can make mistakes. We say that plainly because you should hear it from us. The sandbox and approvals exist so a mistake stays an inconvenience instead of an incident. **What you can do with it** - Research a topic and get a structured report, with web searches routed through Clovy's privacy proxy. - Schedule routines that run on their own, sandboxed by default. - Search and use your own notes, transcripts, and dictation history as context. That data stays on your Mac. - Extend the agent with skills and toolsets you can edit in the app. **Open source, so you can check** Clovy is MIT licensed and the code is public. The agent builds on the open-source Hermes framework, and the backend that routes model calls runs in a TEE with cryptographic attestation. Privacy you can verify beats privacy you are asked to believe. Q: Does Clovy's agent run the AI model on my Mac? A: The agent runs on your Mac: its work, files, memory, and sessions stay on your device. The model that powers its thinking is called over the network, routed to zero-retention models by default, so nothing is stored and nothing is trained on. Q: Can the agent break things on my computer? A: Sessions are sandboxed by default, so writes are blocked outside the agent's workspace and secrets like SSH keys are off limits. Risky actions wait for your approval. The agent can make mistakes, which is exactly why those guardrails are on by default. Q: What is the agent built on? A: Clovy's agent is built on Hermes, the open-source agent framework from Nous Research. Clovy bundles a pinned, verified version and adds private model routing, a sandbox, and deep integration with your notes and meetings. Q: Can the agent run tasks on a schedule? A: Yes. Routines let the agent run tasks on a schedule, and they are sandboxed by default just like regular sessions. Q: Is the agent free to use? A: Clovy's Hobby plan is free and includes light agent usage with the same privacy standard as Pro. Pro is $20 per month and adds more usage, extended agent sessions, skills and toolsets, and scheduled routines. ## FAQ Q: What is Clovy? A: A private AI workspace for the Mac: chat, dictation into any app, meeting notes, and a local agent in one desktop app. Built by OpenSoftware, open source, free to start. Q: Is Clovy free? A: Yes. The Hobby plan is free and includes dictation, meeting notes, and light agent usage. Pro is $20 per month for more usage, extended agent sessions, all models, and scheduled routines. Q: What does Clovy run on? A: macOS 14 or later, on Apple Silicon and Intel. Windows has not shipped yet. Q: Is Clovy open source? A: Yes, MIT licensed, including the backend. The code is on GitHub. Q: Does Clovy work offline? A: Partially. Your notes, transcripts, and files live on your Mac and stay readable offline. Model calls need the internet, so dictation, note generation, and the agent do not work without it. Q: Does Clovy train on my data? A: No. OpenSoftware never trains on your data, and model calls route to zero-retention Venice models by default. Anonymized third-party models are opt-in and labeled; providers on that tier may retain what they receive. Q: What does OpenSoftware store about me? A: Account, login, and billing records. Recordings, transcripts, notes, and agent files are stored on your Mac, not our servers. The exception is anything you deliberately send us, like an issue report. Q: Does anything leave my Mac? A: Yes. Prompts and audio leave for model inference, routed to zero-retention models by default, where nothing is stored or trained on. Your files, history, and agent state stay on your device. Q: How do I verify the privacy claims? A: The backend runs in a TEE and serves its own cryptographic attestation at opensoftware.co/verify. The code is open source, so what we run is what you can read. Q: Can the agent damage my files? A: Sessions are sandboxed by default: writes are blocked outside the agent's workspace, and secrets like SSH keys are on a denylist. Risky actions ask for approval first. The agent can make mistakes; the guardrails assume it will. Q: Where do my prompts go? A: To the Clovy API, Clovy's backend, and from there to the model provider. By default that provider is Venice, under zero-retention terms: nothing stored, nothing trained on. Q: What is Venice? A: A privacy-focused inference provider and Clovy's default. Calls on the Private tier carry contract-enforced zero data retention; the details are on Venice's privacy page. Q: What is the TEE backend? A: The Clovy API runs inside a trusted execution environment, a confidential VM that cryptographically attests to exactly what code it is running. It holds the provider keys and strips identifying metadata from requests, and you can check the live attestation at opensoftware.co/verify. Q: Can I use OpenAI or Claude models in Clovy? A: Yes, if you choose them. The catalog includes OpenAI and Claude models on the Anonymous tier, routed through Venice, and you can opt into OpenAI transcription models, routed from the TEE with the Clovy API as the anonymization layer. Identifying metadata is stripped either way, and Anonymous-tier providers may retain what they receive. Q: What do the Private, Anonymous, and E2EE tiers mean? A: Private is the default: zero retention, contract-enforced. Anonymous models are opt-in, and those providers may retain what they receive. E2EE encrypts your prompt on your Mac, so it is decrypted only inside the enclave that runs the model. Q: Does a bot join my meetings? A: No. Clovy records from your Mac, so nothing joins the call and nothing appears in the participant list. When it detects a supported meeting, it offers to write the notes. Q: How does dictation work? A: Hold the fn key, talk, let go. Clovy types the cleaned-up text wherever your cursor is, in any app, with standard, casual lowercase, and formal styles. There is a hands-free toggle for longer stretches. Q: What can the agent do? A: Research, reports, scheduled routines, and questions over your notes and meetings. It runs on your Mac, built on the open-source Hermes framework, and you can extend it with skills and toolsets in the app. Q: Which AI models does Clovy use? A: Zero-retention Venice models by default, with a catalog you can pick from. Every model in the picker shows its privacy tier: E2EE, Private, or Anonymous. Q: Can I bring my own API key? A: Yes. Set your own Venice API key in settings and Clovy routes your model calls with it instead of metering them through your balance. ## Clovy changelog Latest stable release: v0.0.39 ### v0.0.39: Personalized onboarding and companion controls Published: 2026-07-31T19:20:53Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.39 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.39/June_universal.dmg Clovy now adapts its personality to how you work, expands what you can do from the companion app, and makes interrupted agent runs easier to recover. - Choose how Clovy responds during onboarding or from Agent settings. - Browse approved folders, attach files, and choose models from the companion app. - Review generated media and approve eligible Mac actions from your phone. - Stop agent runs waiting for approval without losing your draft. - See specific failure guidance and report problems directly from Clovy. See more: - Generic assistants rarely match how you prefer to work. Now, choose Clovy's personality during onboarding or change it anytime in Agent settings. - The linked companion app can now browse approved folders, add files to agent runs, choose models, and show generated media. You can also approve eligible Mac actions from your phone when you opt in. - Waiting agent runs now include a Stop control that preserves your draft. Clovy gives clearer guidance when a run fails, restores direct issue reporting, and keeps Linear connections and approvals more reliable. ### v0.0.38: Session archive and calendar details Published: 2026-07-29T04:46:03Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.38 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.38/June_universal.dmg Clovy now keeps archived sessions easy to find, shows matched calendar details in meeting notes, and makes dictation, routines, and connected workspaces more reliable. - Filter sessions by Active, Archived, or All without cluttering the sidebar. - Preview matched events from meeting notes and open them in Google Calendar. - Run routines on demand with connected tools and skills available. - Use connected Notion workspaces and Obsidian vaults with the built-in agent. - Fixed canceled dictation occasionally appearing in the active app. See more: - Finished chats can crowd the sidebar when you still need them later. Now, archive sessions and find them with Active, Archived, or All filters on the Sessions page. - Meeting notes now show matched Google Calendar events in a compact chip. Hover for event details, or open the event in Google Calendar. - Routines now run reliably on demand and keep connected tools available when scheduled. Connected Notion workspaces and Obsidian vaults work with the built-in agent again. Fixed canceled dictation occasionally appearing in the active app. ### v0.0.37: Automatic meeting recording completion Published: 2026-07-28T01:22:07Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.37 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.37/June_universal.dmg Clovy now finishes meeting recordings when your call ends, with a countdown and controls, while dictation, Home, and sidebar feedback stay clearer. - Finish meeting recordings automatically when your call ends. - Use the countdown to stop now or keep recording. - Cancel dictation from its controls or press Escape. - See the recording waveform respond more smoothly to your voice. - Keep Home notices and completed sessions from overlapping controls. See more: - It is easy to leave a recording running after a meeting ends. Now, Clovy detects when your call closes and starts a visible countdown before finishing the recording safely. - You stay in control during the countdown. Stop immediately, keep recording, or let Clovy finish and begin processing the note automatically. - Dictation now has clearer cancel and processing states, including an Escape shortcut. Recording waveforms respond more smoothly, while Home notices and completed sessions stay within their controls. ### v0.0.36: GitHub connections and persistent Home Published: 2026-07-27T17:25:26Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.36 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.36/June_universal.dmg Clovy now connects to GitHub, keeps Home as an ongoing conversation, and gives you faster model control, clearer recording status, and more reliable notes. - Connect GitHub so Clovy can read repositories and help manage issues. - Keep one ongoing Home conversation across restarts and project handoffs. - Search models with /model and choose Instant, Medium, or Hard reasoning. - See recording status in the menu bar while dictating or capturing notes. - Import existing tool connections from Claude, Cursor, or Codex. See more: - Repository work often pulls you away from the conversation. Now, connect GitHub so Clovy can read repositories, search issues, and help manage them with your approval. - Home now stays as one ongoing Clovy conversation, with project handoffs and history that survive restarts. You can search models with /model and choose Instant, Medium, or Hard reasoning for each session. - Live transcripts follow the newest words while you stay near the bottom, and the menu bar shows when Clovy is recording. Fixed notes occasionally losing concurrent edits, generated titles, or older previews. ### v0.0.35: Profiles and connected workspaces Published: 2026-07-21T16:29:30Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.35 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.35/June_universal.dmg Clovy now separates chats and notes into profiles, pulls in more context from your tools, and makes sessions, startup, and transcription recovery easier to trust. - Split Clovy into profiles with isolated chats, notes, and sessions. - Connect Notion and Obsidian, and pull Calendar context into meeting notes. - Mark finished sessions complete and jump between search matches in Files. - Launch Clovy at login and get routine-finished notifications. - Download original note audio and see transcription failures more clearly. See more: - Work gets messy when every chat, note, and recording lives in one stream. Now, Clovy lets you create separate profiles and switch between them, so different projects stay isolated. - Connected context is broader too. Clovy now connects Notion and Obsidian, brings Google Calendar details into meeting notes, and can import Claude Code projects when you want to bring existing work in. - Daily use is steadier as well. You can launch Clovy at login, mark sessions complete, jump between search matches in Files, download original note audio, and see startup or live transcription failures instead of guessing. ### v0.0.34: Private sharing and steadier recovery Published: 2026-07-16T17:45:49Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.34 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.34/June_universal.dmg Clovy now adds private encrypted sharing in the browser, makes model and chat controls easier to trust, and keeps recordings, dictation, and generated media from slipping out of place. - Share Clovy content through private encrypted browser links. - Change models between agent runs and keep Auto easy to reach. - Find generated images in Files and avoid duplicate media blocks. - Open notification clicks in the exact completed chat. - Keep meeting-note transcription and dictation recovery steadier. See more: - Sharing work usually means exporting files or sending people back into the app. Now, Clovy can publish private encrypted shares in the browser so you can send what matters without exposing the rest of your workspace. - Agent chats are easier to control too. You can change models between runs, keep Auto available in the picker, hide internal status chatter, condense long tool runs, and reopen the right completed chat from a notification click. - Capture and recovery feel steadier as well. Meeting-note transcription now survives restarts more reliably, recording controls stay available while notes finish processing, dictation recovers more cleanly after helper interruptions, and generated media stays registered in Files. ### v0.0.33: Private connectors and note exports Published: 2026-07-13T21:16:09Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.33 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.33/June_universal.dmg Clovy keeps connected Google work on your Mac, lets you export meeting notes as PDFs, and makes model picking, dictation, tabs, and project filing easier to trust. - Connect Google tools locally so email and calendar data stay on your Mac. - Export meeting notes as PDFs when you need to share them. - Use clearer Auto model presets instead of stale model picks. - Create projects inline while filing notes or sessions. - Keep long dictation punctuation and issue-report videos intact. See more: - Connected work is more private in this release. Now, Clovy can connect Google tools in local mode so provider tokens stay in macOS Keychain and your email and calendar data do not have to route through Clovy before the agent can use them. - Sharing and organization are easier too. Clovy now exports meeting notes as PDFs, lets you create a project inline while filing a note or session, adds text-size controls and a dedicated Appearance tab, and opens session rename in a clearer dialog. - Everyday work stays steadier as well. Auto is now the suggested model experience, old model picks fall back more cleanly, tabs are easier to reorder and close, long dictation keeps its punctuation better, and issue reports stop dropping extra video attachments. ### v0.0.32: Clearer chats and steadier dictation Published: 2026-07-09T19:50:49Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.32 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.32/June_universal.dmg Clovy names new chats more usefully, renders generated images inline, keeps long dictations aimed at the right app, and makes recording status and notifications easier to trust. - Name new chats from the first exchange or rename them anywhere. - See generated and edited images inline in chat. - Keep long dictations pasting into the app you dictated into. - Detect more meeting and call apps automatically. - See when system audio needs macOS 14.2 or later. See more: - Agent chats are easier to return to when titles explain the work. Now, Clovy names new chats from the first exchange, lets you rename them wherever sessions appear, and shows generated images inline so visual replies stay readable. - Long dictation is steadier too. Clovy keeps the paste target pinned to the app you dictated into, shows when transcription is still running, and stays honest about when system audio capture needs macOS 14.2 or later. - The rest of the app feels calmer as well. Toast notifications replace easy-to-miss chat banners, approval cards are quieter to scan, onboarding asks what you want to use Clovy for, and meeting detection recognizes more call apps. ### v0.0.31: Steadier recordings and cleaner AI settings Published: 2026-07-08T09:38:21Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.31 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.31/June_universal.dmg Clovy keeps recordings and dictation from failing as quietly, lets the agent control visible recordings, and makes AI settings easier to scan when you pick models or manage tools. - Catch silent or stalled recordings before a note gets stranded. - See microphone permission issues before recording fails. - Keep dictation shortcuts from getting stuck after canceled starts. - Let the agent start and stop a visible recording for you. - Browse image models through Suggested and All models lists. See more: - Recording only helps when Clovy captures real audio. Now, Clovy checks microphone access earlier, spots silent captures more clearly, and tears down stalled starts before they leave an empty note behind. - Dictation and agent runs recover more cleanly too. Shortcut starts and stops no longer wedge as easily after lost helper responses, and Clovy shows a clear failure when an agent session stops before replying. - AI setup is easier to scan as well. Clovy cleans up model, skill, server, and external-directory settings, restores image controls to the right place, and hides the full image catalog behind an All models list. ### v0.0.30: Ask Clovy about any note Published: 2026-07-07T18:02:29Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.30 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.30/June_universal.dmg Clovy adds note-aware chat, lets the agent edit attached images more safely, and keeps dictation, transcripts, and issue reports steadier while you work. - Open a note-specific Ask Clovy chat from any note. - Reference specific notes directly in agent chat. - Edit attached images without leaving the conversation. - Keep dictation from bleeding into live meeting recordings. - Paste screenshots straight into issue reports. See more: - Meeting notes are more useful when Clovy can answer questions in context. Now, every note gets its own Ask Clovy panel, and you can reference specific notes directly in chat without copying their contents by hand. - Image work is smoother too. Clovy now lets the agent edit attached images in place, and safe mode starts on by default so image generation stays filtered unless you choose otherwise. - Capture and troubleshooting are steadier across the app as well. Dictation no longer spills into live recordings, transcript loss is easier to spot, and issue reports accept pasted screenshots when you need to show exactly what went wrong. ### v0.0.29: Smarter dictation and clearer upgrades Published: 2026-07-03T15:13:57Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.29 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.29/June_universal.dmg Clovy formats dictated email more naturally, makes plan changes easier to follow, and stays more usable when sign-in, transcripts, or chat controls hit rough edges. - Format dictated text like an email in mail apps. - Use a clearer session usage panel and plan upgrade flow. - Keep sign-in working through auth-service outages. - See steadier live transcript preview for system audio. - Get cleaner chat actions, tooltips, and composer scrolling. See more: - Dictation is easier to trust when it matches where you are writing. Now, Clovy can lay out dictated text like an email in mail apps while keeping live transcript preview steadier for system audio. - Billing is clearer too. Clovy now shows a redesigned session usage panel, adds a calmer Upgrade to Max flow, and keeps the plan-refresh path steadier after you confirm a change. - The rest of the app stays out of your way more consistently as well. Sign-in can recover through auth-service outages, chat actions and tooltips feel less cramped, and composer scrolling holds together better while you draft. ### v0.0.28: Steadier dictation and recovery Published: 2026-07-02T15:48:00Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.28 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.28/June_universal.dmg Clovy keeps dictation working through permission gaps and transient outages, preserves your wording more faithfully, and explains blocked actions or retries instead of failing silently. - See the Accessibility prompt before dictation paste fails. - Keep dictation available when the helper needs a relaunch. - Keep your original wording when Clovy cleans up dictation. - Recover more cleanly from transient sign-in and transcription outages. - See clearer reasons when branching is unavailable or a retry can work. See more: - Dictation is only useful when it shows up where you need it. Now, Clovy surfaces Accessibility permission issues before paste fails, leaves the text ready for manual paste when needed, and relaunches the dictation helper if it unexpectedly drops out. - Cleanup is more faithful too. Clovy now preserves your original wording more closely, keeps dictated numbered lists intact, and carries language information forward more reliably when transcription comes back incomplete. - Recovery is steadier across the rest of the app as well. Temporary account or transcription outages stay retryable instead of knocking you out of a session, and Clovy explains blocked branching or upstream errors instead of leaving you guessing. ### v0.0.27: Local models and steadier chats Published: 2026-07-02T04:39:32Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.27 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.27/June_universal.dmg Clovy lets you run supported local text models, keeps long meeting transcripts from dropping system audio, and makes chat actions and scrolling cleaner to use. - Run supported local text models inside Clovy. - Keep system-audio transcripts through long or quiet meetings. - Use cleaner message actions with a jump-to-latest pill. - Choose from refreshed clay-toned themes and icons. - Pick a Max upgrade path when you need more headroom. See more: - Some work needs local inference. Now, Clovy can connect to supported local text models, and advanced model settings stay tucked away until you need them. - Long chats are easier to manage too. Message actions are cleaner, branching is more reliable, and a jump-to-latest control helps you catch back up after scrolling away from the newest reply. - Meeting notes hold together better in the background as well. Clovy now keeps system-audio transcription from dropping out during long or quiet meetings, and billing settings give you a clearer way to move up to Max. ### v0.0.26: Bring your own Venice key Published: 2026-07-01T21:14:14Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.26 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.26/June_universal.dmg Clovy lets you add your own Venice API key, warns before oversized sends, and keeps drafts, meeting notes, and system-audio transcripts from slipping away. - Add your own Venice API key in the app. - See a warning before a prompt is too large to send. - Keep new-session drafts from disappearing. - Keep meeting notes after deleting the last note. - Restore missing system-audio transcription. See more: - Some setups need more control over model access. Now, you can add your own Venice API key directly in Clovy and keep working with the account setup you prefer. - Sending is safer too. Clovy now warns before an oversized prompt goes out, preserves drafts when you start a new session, and keeps the active-session report action ready when you need to share what happened. - Voice and account recovery are steadier as well. Meeting notes stay put after you delete the last note, system-audio transcription comes back when the source drops out, and you can sign out of OS Accounts without leaving the app. ### v0.0.25: Image generation and release channels Published: 2026-07-01T17:12:56Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.25 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.25/June_universal.dmg Clovy adds built-in image generation, lets you switch updater channels, adds accent themes, and recovers more cleanly when transcripts, permissions, or long prompts go sideways. - Generate images directly inside Clovy. - Switch the updater between stable and RC channels. - Choose from six accent themes with matching dock and HUD styling. - Recover from context-length overflows instead of getting stuck. - Get better meeting-note recovery, transcript evidence, and issue-report uploads. See more: - Clovy now covers more creative work without leaving the app. You can generate images directly in Clovy, and the updater can move between stable and release-candidate channels without making you juggle separate installs. - Recovery is much stronger too. Failed transcript evidence is easier to inspect, meeting notes survive more long-turn failures, issue-report attachments publish more reliably, and context-length overflows turn into something you can recover from instead of a dead stop. - The app also feels more personal day to day. Six accent themes, themed dock and HUD styling, dismissible dictation permission prompts, clearer account labels, and better accessibility-permission refreshes make the desktop easier to trust. ### v0.0.24: Cleaner billing and calmer settings Published: 2026-06-30T13:28:48Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.24 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.24/June_0.0.24_universal.dmg Clovy makes plan status and remaining usage easier to read, treats paid accounts more consistently, and removes distracting native context menus in production. - See billing usage as remaining percentage. - Read a clearer plan header and usage meter. - Get cleaner Free plan and paid-plan labels. - Avoid native context menus in production. - See a more polished Clovy logo. See more: - Billing is easier to scan in this release. Clovy now shows a clearer plan header, a more legible usage meter, and labels that explain free and paid states without making you decode them. - Paid accounts also read more consistently. When subscription data comes in incomplete, Clovy is better at treating subscribed users like subscribed users instead of showing a confusing in-between state. - A few visual edges are calmer too. Native context menus stay out of production, and the Clovy logo gradients look more intentional across the desktop app. ### v0.0.23: Long recordings and clearer image prompts Published: 2026-06-29T21:18:57Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.23 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.23/June_0.0.23_universal.dmg Clovy handles longer recordings more reliably, refreshes system-audio permissions better, and explains when a model cannot work with an image. - Keep longer recording audio moving through processing. - Refresh system-audio permission state more reliably. - See a clear banner when the selected model cannot use images. - Get cleaner transcript failure feedback. See more: - Longer recordings only help if they finish cleanly. Now, Clovy keeps more recording audio moving through processing instead of stalling before the transcript is ready. - Image attachments are easier to recover too. Clovy now explains when the model you picked cannot work with images, so you can switch models instead of guessing why the composer stopped helping. - Permission and transcript feedback are cleaner as well. Settings stays closer to the real macOS audio state, and transcript failures are easier to understand when something upstream breaks. ### v0.0.22: Session restore and clearer recovery Published: 2026-06-29T13:07:30Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.22 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.22/June_0.0.22_universal.dmg Clovy restores agent tabs more consistently, keeps subagent views easier to follow, and gives subscribers clearer next steps when account balance blocks work. - Restore agent tab sessions more reliably. - Follow subagent views without jumpy scrolling. - See clearer recovery guidance when balance blocks work. - Get steadier account-state messaging. See more: - Losing your place after a restart breaks trust quickly. Now, Clovy restores agent tabs more reliably so you can get back to the session you were already using. - Subagent views are easier to follow too. Scrolling is steadier, which helps longer runs feel less chaotic when several threads are active at once. - When account balance gets in the way, Clovy points subscribers toward the right recovery path with clearer guidance instead of leaving the block vague. ### v0.0.21: Safer web browsing defaults Published: 2026-06-27T19:29:06Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.21 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.21/June_0.0.21_universal.dmg Clovy tightens agent web-browsing protections so automated fetches stay away from private or unsafe network targets while normal public lookups keep working. - Check agent web addresses before browsing starts. - Keep automated fetches away from private or unsafe network targets. - Preserve normal public website fetches after stricter checks. - Keep the command palette clear above desktop tabs. See more: - Agent web tools are more useful when they are safe by default. Now, Clovy validates browsing targets before the agent fetches them, keeping automated requests away from private network addresses and other risky destinations. - The stricter checks do not make normal browsing harder. Public website fetches still work as expected, so the added protection stays out of your way when you are using Clovy for everyday research. - A smaller desktop fix keeps the command palette visually clear above tab chrome when you navigate Clovy from the keyboard. ### v0.0.20: Web tools and agent controls Published: 2026-06-26T19:39:41Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.20 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.20/June_0.0.20_universal.dmg Clovy adds web search and fetch, composer slash commands, better message controls, and a broad set of fixes across dictation, sessions, and account recovery. - Search the web and fetch pages from inside Clovy. - Use slash commands, copy, and edit on agent messages. - Get steadier session ordering, branching, and queued steering. - See cleaner recovery for dictation and account blocks. See more: - Clovy can do more without breaking your flow in this release. You can now search the web, fetch pages, use slash commands, and copy or edit agent messages without leaving an active session. - Longer runs are easier to manage too. Session ordering, branching, queued steering, drafts, model switching, and tool activity all behave more predictably while the agent is still working. - Recovery improves across the edges as well. Dictation fallback, push-to-talk after hands-free mode, note-processing progress, checkout routing, and the guidance shown in Settings all get easier to work through. ### v0.0.19: Recording recovery and steadier retries Published: 2026-06-18T16:45:10Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.19 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.19/June_0.0.19_universal.dmg Clovy recovers interrupted meeting recordings better, pauses stale recordings cleanly, and makes transcription retries more resilient. - Recover interrupted paused recordings. - Pause stale meeting recordings cleanly. - Retry transcription more reliably. - Keep agent error banners tied to the right session. See more: - Meeting recording is steadier here. Clovy now recovers better after interruptions, pauses stale recordings cleanly, and gives transcription retries a better chance to finish instead of leaving a recording half-alive. - Agent errors are scoped more cleanly too, so one session is less likely to make the rest of the app feel broken. Bug submission also holds together better when a downstream delivery step has trouble. ### v0.0.18: Scoped note failures Published: 2026-06-17T21:37:14Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.18 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.18/June_0.0.18_universal.dmg Clovy keeps transcription failures tied to the note that failed and sends issue reports to the right destination more reliably. - Keep a failed transcript tied to the note that failed. - Prevent one broken note from making other notes look broken. - Send issue reports to the right destination. See more: - One broken note should not poison the rest of your workspace. Now, Clovy keeps transcript failures scoped to the note that actually broke instead of spreading that state across unrelated notes. - Issue reports also land in the right destination more reliably, so feedback and bug reports are easier to follow up on after you send them. ### v0.0.17: Manual update checks Published: 2026-06-17T19:07:49Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.17 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.17/June_0.0.17_universal.dmg Clovy adds a manual update check in About settings, improves issue-report delivery, and fixes a startup hang in the messaging surface. - Check for updates from About settings. - Send issue reports more reliably. - Avoid a startup hang in the messaging surface. - Edit agent skills in a roomier panel. - Get steadier sign-in actions in account settings. See more: - Waiting for a background update prompt is no longer your only option. Clovy now gives you a direct update check in About settings so you can verify availability on demand. - Issue-report and messaging fixes remove two frustrating blockers at once. Reports are more likely to reach the right place, and the messaging surface is less likely to hang while it starts up. ### v0.0.16: Live transcripts and skill editing Published: 2026-06-17T16:24:37Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.16 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.16/June_0.0.16_universal.dmg Clovy adds live transcript preview, editable agent skills, active routine history, and steadier behavior across note tabs and desktop chrome. - Watch live transcript preview while recordings are active. - Edit agent skills inside Clovy. - See active routine runs in history. - Get cleaner note retry and live transcript states. - See steadier dialogs, note tabs, and tab chrome. See more: - Clovy is easier to work from moment to moment in this release. You can watch transcription progress live, edit agent skills without leaving the app, and see active routine runs where they already belong. - Supporting fixes smooth out the rest of the workflow. Note retry states, dialogs behind tabs, referral overflow, tab chrome clipping, auth callbacks, and file filtering all behave more cleanly. ### v0.0.15: Tabs and faster dictation Published: 2026-06-16T20:02:51Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.15 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.15/June_0.0.15_universal.dmg Clovy adds browser-style tabs, in-app referral links, periodic update checks, and a faster, cleaner dictation paste path. - Open work in a browser-style tab bar. - Use Command-N for new agent sessions. - Send bug, feedback, and feature requests from the composer. - Filter filler words and speed up dictation paste. - Get in-app referral links and periodic update checks. See more: - Clovy feels much more like a desktop workspace here. Tabs, draggable tab chrome, a clearer Sessions nav, and contained update prompts make it easier to keep several threads of work open at once. - Dictation also gets faster and cleaner. Filler words are filtered out more intelligently, paste happens sooner, processing errors stay better scoped, and generated notes are less likely to inherit messy source labels. ### v0.0.14: Startup recovery and native settings Published: 2026-06-15T21:37:01Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.14 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.14/June_0.0.14_universal.dmg Clovy restores sessions more reliably at launch and adds native macOS Settings entries so desktop controls feel more at home on your Mac. - Retry session restore at startup. - Open Settings from native macOS menu entries. - See desktop controls move closer to Mac conventions. See more: - Reopening Clovy is less likely to strand you in an empty state now. Session restore retries at startup, which gives returning work a better chance to come back without manual cleanup. - Settings also feel more native. macOS menu entries make the app behave more like a normal Mac desktop tool instead of hiding important controls behind custom navigation. ### v0.0.13: Universal downloads and cleaner updates Published: 2026-06-15T17:09:10Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.13 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.13/June_0.0.13_universal.dmg Clovy starts shipping universal Mac downloads, improves relaunch after updates, and turns model-response failures into clearer app behavior. - Download one build for Apple Silicon and Intel Macs. - Relaunch more smoothly after updates. - See clearer handling when a model response breaks. - Get Kimi K2.6 as the default text model. - Avoid guesswork about which DMG to install. See more: - Installing Clovy is simpler in this release. Universal Mac builds make the download path clearer for both Apple Silicon and Intel machines, so fewer people need to guess which DMG they need. - Updates feel cleaner too. Clovy relaunches more reliably after install, and broken model responses show up as clearer app behavior instead of spilling raw failures into the interface. ### v0.0.12: Routines and instant stops Published: 2026-06-13T14:22:33Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.12 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.12/June_0.0.12_aarch64.dmg Clovy adds routines management, smooths dictation HUD transitions, and makes stopping an agent session feel immediate. - Manage routines inside Clovy. - Stop agent sessions instantly in the UI. - See smoother dictation HUD transitions. See more: - Routines are easier to work with once they live inside the app. Clovy now gives you direct management controls and makes stopping a running agent feel immediate instead of lagging behind your click. - Voice controls feel calmer too. The dictation HUD transitions more smoothly, so starting and stopping speech work looks less jumpy while you are in flow. ### v0.0.11: Menu bar polish and audio setup Published: 2026-06-13T03:15:47Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.11 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.11/June_0.0.11_aarch64.dmg Clovy polishes the menu bar and HUD, moves model controls into the composer, and asks for system-audio permission during onboarding. - Move model controls into the composer. - See new sessions marked active in the sidebar. - Get a more polished menu bar and agent HUD. - Grant system-audio permission during onboarding. - See delegated subagents inside agent chat. See more: - Everyday navigation is tighter here. Model controls move closer to where you type, empty states are clearer, model descriptions are easier to inspect, and the menu bar plus HUD feel more finished. - Meeting setup is clearer too. Clovy asks for system-audio permission during onboarding instead of waiting until you expect call audio to already be working. ### v0.0.10: Faster meetings and pinned chats Published: 2026-06-12T18:53:40Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.10 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.10/June_0.0.10_aarch64.dmg Clovy speeds up meeting processing, adds pinned agent sessions, and makes the menu bar feel more native. - Stop reprocessing the whole recording for each meeting turn. - Pin agent sessions in the sidebar. - See the Clovy mark in the menu bar. - Get a cleaner already-listening cue. See more: - Longer meetings move faster in this release. Clovy no longer has to reprocess the entire recording every time a new meeting turn arrives, which keeps active calls from feeling needlessly heavy. - Pinned sessions and the menu bar mark make the app easier to come back to during repeated agent workflows, and the listening state is easier to understand at a glance. ### v0.0.9: Simpler shortcuts and cleaner recovery Published: 2026-06-12T16:44:43Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.9 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.9/June_0.0.9_aarch64.dmg Clovy makes shortcuts work with fewer permissions and helps agent sessions recover more cleanly from context overflow or missing access. - Use shortcuts without Input Monitoring. - Approve Agent CLI access directly in chat. - Recover from prompts that run past the model limit. - See and switch the current model from the session bar. - Get clearer routine history and session-mode cues. See more: - Shortcuts are easier to trust when they ask less from your Mac. Clovy now handles that path without Input Monitoring, which lowers setup friction for keyboard-first use. - Agent sessions recover more cleanly too. Prompts that run past the model limit turn into something the agent can work with, CLI access can be approved in chat, and the current model stays visible in the session bar. - Routine history and session-mode cues are clearer as well, so it is easier to tell what Clovy is doing and how much access a session is using. ### v0.0.8: Session continuity and smoother starts Published: 2026-06-11T18:15:03Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.8 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.8/June_0.0.8_aarch64.dmg Clovy keeps active agent sessions alive during app refreshes, adds stable Mac download links, and makes trial starts and session transitions feel smoother. - Keep in-flight agent sessions alive during app refreshes. - Get stable Mac download links. - Open verify links through the desktop app. - Pick models from a curated Suggested tab. - See the agent HUD take over from the mascot. See more: - Long-running agent work is less fragile here. Clovy keeps active sessions alive during app refreshes, so you are less likely to lose progress in the middle of a task. - Starting a trial feels quicker too. Checkout opens with fewer awkward pauses, which makes first-run account flows feel less brittle. - The rest of the app catches up around that. Session transitions are smoother, model picking starts with better suggestions, sidebar states are clearer, and Mac download links are easier to rely on. ### v0.0.7: Built-in agent and bug reports Published: 2026-06-11T16:52:19Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.7 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.7/June_0.0.7_aarch64.dmg Clovy brings the built-in agent deeper into the app, adds bug reports with diagnostics, and keeps session-level access boundaries easier to understand. - Keep the built-in agent inside the app bundle. - Stop stale fetches from erasing chat messages. - Send bug reports with Clovy's diagnosis. - Hide text-only models from agent selection. - Keep broader-access mode scoped to one session. See more: - The built-in agent feels much more native here. It starts with the app, stays closer to your sessions, and asks you to manage fewer separate moving pieces while you work. - Access boundaries are clearer too. Broader-access mode stays scoped to the session that asked for it, which makes mixed workflows easier to understand when some tasks need more freedom than others. - Bug reports become more useful as well. Diagnostics and attachments make reports easier to act on, and a chat fix stops stale fetches from wiping messages you still need. ### v0.0.6: Safer sessions and easier trials Published: 2026-06-11T15:11:02Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.6 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.6/June_0.0.6_aarch64.dmg Clovy makes agent sessions safer by default, improves trial checkout, and adds clearer privacy and scheduling cues across the app. - Start new agent sessions with safer defaults. - Opt a session into fuller access when needed. - Begin sign-in and free-trial checkout from onboarding. - See clearer model privacy cues. - Read routine schedules in plain language. See more: - New agent sessions start with safer defaults now. Clovy keeps a tighter boundary around fresh work while still letting you opt into fuller access when a task genuinely needs it. - Trials are easier to start too. Sign-in and free-trial checkout move into onboarding, and returning to the app after account steps is more reliable. - The rest of the privacy story is clearer as well. Model labels are easier to read, verify guidance is easier to find, and routine schedules show up in plain language instead of a technical format. ### v0.0.5: Clearer access recovery Published: 2026-06-10T19:24:01Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.5 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.5/June_0.0.5_aarch64.dmg Clovy checks trial and subscription access more consistently so sign-in, re-entry, and upgrade paths feel clearer when your account needs attention. - Keep app access aligned with your trial or subscription status. - Make expired-access states less confusing after sign-in. - Use a more reliable recovery path when access needs attention. See more: - Account access is more consistent in this release. Clovy checks trial and subscription status more cleanly, so the app opens in the state your account actually supports instead of making you decode mixed signals. - Recovery is steadier too. When access needs attention, Clovy is less likely to bounce you through the wrong screen before showing the next step that will actually help. ### v0.0.4: Meeting controls and desktop polish Published: 2026-06-10T18:22:01Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.4 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.4/June_0.0.4_aarch64.dmg Clovy introduces meeting detection, a floating recording HUD, a command palette, and clearer settings, permissions, and agent controls across the desktop app. - Detect meetings and record from a floating HUD. - Search with a command palette and dedicated permissions tab. - Test your mic and reset dictation shortcuts. - Track Clovy from a menu bar status item. - Find and download workspace files more easily in agent chat. See more: - Clovy starts feeling like a real desktop workspace here. Meeting detection arrives with a floating recording HUD, and the command palette, settings, and permissions surfaces become easier to navigate. - Voice setup gets more practical too. You can test microphone playback, reset dictation shortcuts, choose language settings, and handle recording consent at a clearer moment in the workflow. - Agent sessions get the same polish. Titles, status, file download cards, approval explanations, and the menu bar status item all make the app easier to monitor while work is in progress. ### v0.0.3: Agent workspace and saved chats Published: 2026-06-05T15:59:07Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.3 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.3/OS.Scribe_0.0.3_aarch64.dmg OS Scribe adds its first built-in agent workspace with saved chats, approval prompts, workspace file access, and steadier recording support. - Open the first built-in agent workspace. - Keep agent conversations saved between sessions. - Browse workspace files tied to an agent session. - See agent approval requests inside the app. - Run the agent through an app-managed setup. See more: - OS Scribe gets its first practical agent workspace in this release. Chats stay saved, approval prompts are visible, activity status is easier to follow, and replies are better organized inside the app. - Session context is easier to inspect too. You can browse the files tied to an agent session without leaving the workspace, which makes generated artifacts easier to track. - Recording reliability improves alongside that new agent surface, with better recovery from stuck recordings and steadier long-running capture sessions. ### v0.0.2: Auto-updates and steadier recording Published: 2026-06-04T13:33:57Z Release: https://github.com/open-software-network/os-june-releases/releases/tag/v0.0.2 Download: https://github.com/open-software-network/os-june-releases/releases/download/v0.0.2/OS.Scribe_0.0.2_aarch64.dmg The early OS Scribe build adds in-app updates, steadier signed Mac builds, better recording recovery, and a more polished dictation HUD. - Install macOS updates from inside the app. - Get a steadier recording queue and audio path. - See incremental turn transcription and better recorder waveforms. - Use a more polished dictation HUD. - Run more reliable signed Mac builds with audio support. See more: - Keeping OS Scribe current gets much easier here. In-app updates arrive early, and signed Mac builds are more reliable, so staying current feels less manual. - Voice gets a much stronger base at the same time. Recording recovery, incremental transcription, waveform behavior, and the recorder UI all improve together, which makes early dictation and meeting capture feel steadier. ## Published news # Introducing Clovy > Clovy is a private, local-first AI workspace for Mac with chat, dictation, meeting notes, and an open-source local agent. - Canonical URL: https://www.opensoftware.co/news/introducing-clovy - Published: Jun 29, 2026 - Updated: Jun 29, 2026 - Category: Product - Author: OpenSoftware - Reading time: 3 min read - Topics: Clovy, private AI, local-first AI, Mac AI assistant, AI agent, voice dictation, meeting notes, open source AI, Venice AI, Hermes Agent, Clovy community, OpenSoftware ## Primary links - [Learn more about Clovy](https://www.opensoftware.co/clovy) - [Download Clovy](https://www.opensoftware.co/download/mac) - [Clovy community](https://t.me/+B4Z8KUqEsRc4ZGVh) - [Open source repository](https://github.com/open-software-network/os-clovy) - [Venice AI](https://venice.ai/) - [Hermes Agent](https://hermes-agent.nousresearch.com/) - [Hermes Agent source](https://github.com/NousResearch/hermes-agent) Clovy is a private way to use AI and agents on your Mac. It bundles the AI workflows people already use into a single private workspace: a local agent and chat, voice dictation into any app, automated meeting notes, private open-source models, and anonymized frontier models. It is built so your files, context, usage, and data stay yours. Clovy is fully open source, local by default, and we do not train models on your data. You can [learn more about Clovy](/clovy), [download Clovy](/download/mac), or inspect the [open-source code on GitHub](https://github.com/open-software-network/os-clovy). We are handing AI some of our most precious and sensitive data: our thoughts, our work, and access to our day-to-day lives. Clovy was built so that access to powerful AI does not come at the expense of privacy. Files, transcripts, sessions, memory, and agent state stay local by default. For inference, Clovy defaults to private [Venice AI](https://venice.ai/) models designed not to retain your prompts, audio, transcripts, or data. ## Working with Clovy Clovy is designed around the AI workflows people already use on a desktop. Chat is for questions, planning, research, and quick thinking. It gives you one place to work with models while keeping the surrounding product quiet. Dictation lets you press a shortcut, speak naturally, and have Clovy type into the app you are currently using. Your emails, chats, notes, browser, documents, and anywhere else your cursor is. Your thoughts are no longer constrained by the speed of typing on a keyboard. Meeting notes help capture what happened while you focus on the call. Clovy can detect supported meetings and be triggered to take notes without a bot joining the call and sitting awkwardly in the room. Decisions, action items, follow-ups, transcripts, and notes are stored locally after private processing. The agent is for longer-running work. Clovy comes with a local agent built on the open-source [Hermes Agent](https://hermes-agent.nousresearch.com/). Give it a task, and it works through it on your machine with the tools and files you give it access to. It can draft a document, pull together research, prepare a report, and schedule recurring routines. It learns from you and becomes more useful over time. The agent defaults to sandboxed mode, and context, sessions, chats, and runtime stay local to your device. ## Getting started Clovy is available today for macOS. Free to try, no card required. [Learn more](/clovy), [download the latest Mac build](/download/mac), or join the [Clovy community](https://t.me/+B4Z8KUqEsRc4ZGVh). The code is open source, so developers can inspect how the product works, track what is changing, or contribute improvements in the [OpenSoftware Clovy repository](https://github.com/open-software-network/os-clovy). ## Private by default Clovy keeps files, sessions, memory, and agent state local by default. Model calls are routed through privacy-preserving inference, with private open-source models as the default path and anonymized frontier model routing available when you need it. ## Built in the open The Clovy codebase is open source so people can inspect the product instead of only trusting the website. The backend is designed to run with [verifiable infrastructure](/verify), and the product stores only the account, login, and billing records needed to operate the service. ## Blog guides # The best ChatGPT desktop alternatives for Mac in 2026 > Seven ChatGPT desktop alternatives for Mac compared on privacy, local models, voice input, file access, automation, hardware support, and price. - Canonical URL: https://www.opensoftware.co/blog/chatgpt-desktop-alternatives-for-mac - Published: Jul 20, 2026 - Updated: Jul 20, 2026 - Category: Guides - Author: OpenSoftware - Reading time: 9 min read - Topics: ChatGPT desktop alternatives Mac, ChatGPT Mac alternative, ChatGPT desktop replacement, local ChatGPT Mac, ChatGPT app alternative macOS, Mac ChatGPT alternative ## Primary links - [Learn more about Clovy](https://www.opensoftware.co/clovy) - [Download Clovy](https://www.opensoftware.co/download/mac) - [Clovy community](https://t.me/+B4Z8KUqEsRc4ZGVh) - [Open source repository](https://github.com/open-software-network/os-clovy) - [Venice AI](https://venice.ai/) - [Hermes Agent](https://hermes-agent.nousresearch.com/) - [Hermes Agent source](https://github.com/NousResearch/hermes-agent) The ChatGPT desktop app is no longer just a smaller version of the website. On Mac it can work with supported apps, upload files and screenshots, and [record meetings](https://help.openai.com/en/articles/11487532-chatgpt-record). It is still a cloud-first product. Files uploaded through the macOS app are stored in OpenAI's cloud, consumer content may be used to improve models unless you opt out, and the app does not run its frontier models on your Mac. The right alternative depends on what you are replacing: general chat, local document analysis, voice input, meeting notes, web research, or desktop automation. This page is narrowly about the Mac desktop experience. For retention and company governance, use the broader [private ChatGPT alternatives for work](https://opensoftware.co/blog/private-alternatives-to-chatgpt-for-work) guide. For a Claude-focused comparison, see [private alternatives to Claude](https://opensoftware.co/blog/private-alternatives-to-claude). ## How we evaluated the Mac apps This is a documentation-based comparison, checked on July 20, 2026. We evaluated native Mac access, inference location, saved-data location, voice and meeting features, file and app access, hardware support, price, and the clarity of each vendor's current documentation. It is not a hands-on model-quality benchmark. ## Quick picks | Alternative | Best for | Native Mac app | Local inference | Works on Intel | |---|---|---:|---:|---:| | [Clovy](https://opensoftware.co/clovy) | Private chat, dictation, meetings, and agent work together | Yes | Optional for text generation | Yes | | [Claude Desktop](https://claude.com/download) | Frontier reasoning, coding, and Cowork | Yes | No | Yes | | [Gemini for Mac](https://gemini.google/mac) | Google-connected work and quick visual context | Yes | No | No | | [LM Studio](https://lmstudio.ai) | Local chat, documents, and a local API | Yes | Yes | Check the current download before installing | | [Jan](https://jan.ai) | Open-source local ChatGPT-style experience | Yes | Yes | No | | [AnythingLLM Desktop](https://anythingllm.com/desktop) | Local document collections and workspaces | Yes | With a local model | Yes | | [Brave Leo](https://brave.com/leo/) | Private browser-based assistance | Through Brave | Optional with a local model | Yes | ChatGPT itself currently supports macOS 14 or later on both Apple Silicon and Intel. Hardware support is therefore a reason to inspect alternatives carefully, not a reason to assume ChatGPT is unavailable. ## What to compare on a Mac Desktop AI products often blur four separate capabilities: 1. **Native access:** a keyboard shortcut, menu-bar surface, screenshots, or app context. 2. **Local storage:** saved conversations and files remain on the Mac. 3. **Local inference:** the model itself runs on the Mac. 4. **Computer action:** the product can use files, applications, browsers, or shell tools with your permission. A desktop app can have native access while sending every prompt to the cloud. A local model can keep prompts on-device while offering none of the polished meeting or voice features of a cloud app. ## 1. Clovy: best integrated private alternative for Mac Clovy is the private way to use AI on your desktop. It combines chat, system-wide push-to-talk dictation, meeting notes without a participant bot, and a local Hermes-based agent. The app and agent runtime run on your Mac. Recordings, transcripts, notes, files, sessions, and memory are stored locally by default. Selected audio, text, files, or context can still leave the Mac when required for inference. By default those requests route through the Clovy API to Venice Private models under zero-retention and no-training terms. Clovy's [verification page](https://opensoftware.co/verify) exposes evidence for the open-source Clovy API running in an attested confidential virtual machine. The attestation covers that API layer, not every upstream model. Clovy labels models by privacy mode so you can distinguish default Private models from optional Anonymous providers that may retain submitted content. For text generation and agent work, Clovy can connect to an OpenAI-compatible local model endpoint. That is optional. Dictation and meeting-note generation are not fully offline by default. Clovy's default content path is zero retention, but the service still keeps account, billing, capacity, model-selection, request-timing, credit, and error records. Issue reports retain what you deliberately submit, and connected tools or websites create separate data boundaries. The app supports macOS 14 or later on Apple Silicon and Intel. Meeting capture with system audio needs macOS 14.2 or later. On macOS 14.0 and 14.1, recording falls back to the microphone. **Better than ChatGPT desktop for:** local-first workspace history, system-wide dictation, automatic meeting notes, scheduled routines, and verifiable private routing. **ChatGPT remains better for:** OpenAI's own frontier models, its broader cloud ecosystem, and teams already standardized on ChatGPT Business or Enterprise. **Price:** Free, Pro at $20/month, and Max at $100/month. Scheduled routines require Pro or above. ## 2. Claude Desktop: best for frontier reasoning and agentic knowledge work Claude Desktop is the closest direct competitor to the current ChatGPT desktop experience. Quick Entry opens from anywhere on macOS, accepts screenshots and window context, and supports voice dictation on macOS 14 or later. Claude Cowork can plan and run longer knowledge-work tasks, use local files through the desktop app, and continue remote tasks on Anthropic's servers. That last point is the privacy boundary. Desktop access does not mean local inference. Cowork cloud sessions run remotely, and saved sessions and files live in the Claude account. For consumer Free, Pro, and Max accounts, Anthropic uses chats for model improvement only if you allow it, with safety and feedback exceptions. Saved chats remain until you delete them. Deleted chats are scheduled for backend deletion within 30 days. Eligible data shared for model improvement may remain in training pipelines for up to five years. Claude for Work does not train on organizational content by default, but saved cloud conversations still follow the workspace's retention rules. Claude Desktop offers a universal Mac build for Intel and Apple Silicon. **Better than ChatGPT desktop for:** people who prefer Claude's writing, reasoning, coding, and Cowork workflows. **Not a privacy-first replacement if:** your requirement is local inference, local-only history, or zero-retention consumer chat. ## 3. Gemini for Mac: best for Google-centered work Gemini's Mac app provides quick entry, window sharing, and access to Google's assistant from the desktop. It requires macOS 15 or later, Apple Silicon, at least 8 GB of memory, and an internet connection. Consumer privacy depends heavily on Gemini Apps Activity. With activity on, Google may use chats to improve services and models, including human review, with an 18-month default auto-delete period. Turning activity off prevents future chats from being used for model training unless you submit feedback, but Google still keeps them for up to 72 hours to provide and protect the service. Qualifying Google Workspace plans have different terms. Workspace content is not used to train generative models outside the Workspace domain without permission. **Better than ChatGPT desktop for:** users who prefer Gemini and already work inside Google's account ecosystem. **Not the best fit for:** Intel Macs, offline work, or users who want local history and local models. ## 4. LM Studio: best for a polished local AI workspace LM Studio runs downloaded models on your Mac and can work entirely offline after the model files and runtimes are installed. Local chat, document retrieval, and its OpenAI-compatible server keep prompts and files on the machine. This creates the clearest technical answer to cloud retention: there is no cloud request to retain. It also moves the burden of model choice, disk space, speed, and quality onto you. The local model must fit in your Mac's available memory, and smaller models may be weaker than hosted frontier models. Optional cloud models and web search use transient processing under zero-retention terms. Turning those on changes the boundary, even though you remain inside LM Studio. **Better than ChatGPT desktop for:** offline conversations, private document analysis, local development, and serving a model to other Mac apps. **Not better for:** automatic meeting notes, system-wide dictation, or users who want frontier quality without thinking about hardware. ## 5. Jan: best open-source local ChatGPT replacement Jan is an Apache 2.0 licensed desktop AI platform. It runs local models, stores its data in a local Jan folder, works without an account, and can operate offline after model download. Optional analytics are consent-based. It also connects to cloud providers and external tools. If you select a cloud GPT or Claude model, those providers receive the request and their policies apply. Jan for Mac requires macOS 13.6 or later and Apple Silicon. The official guide suggests roughly 8 GB for very small models, 16 GB for many 7B-class models, and more memory for larger models or context windows. **Better than ChatGPT desktop for:** open-source ownership, no-account local chat, and experimenting with open models. **Not better for:** Intel Macs, polished voice and meeting workflows, or users who do not want to manage models. ## 6. AnythingLLM Desktop: best for local document collections AnythingLLM Desktop focuses on workspaces built around your documents. It can keep the model, vector database, documents, and chats on one Mac when every component is configured locally and networked tools are disabled. Native downloads are available for Apple Silicon and Intel Macs. Its flexibility is also its main caveat. AnythingLLM can connect to hosted model providers, remote vector databases, and external tools. Privacy depends on the components you select. "Self-hosted" describes deployment, not the behavior of every connected service. **Better than ChatGPT desktop for:** organizing multiple local document collections and choosing each part of the retrieval stack. **Not better for:** a minimal out-of-box setup, system-wide dictation, or meeting capture. ## 7. Brave Leo: best private assistant inside the browser Leo is built into the Brave browser rather than shipped as a separate desktop assistant. Free use does not require an account. Brave says prompts and responses are not retained on its servers or used for training, and conversation history stays locally in the browser. Large prompts may be cached briefly for performance. Leo is useful when the work already happens in tabs. It can summarize pages and help with browser context without becoming a general Mac workspace. Brave also supports connecting selected local or remote models, which changes the available features and privacy boundary. **Better than ChatGPT desktop for:** private page summarization and browser-first work. **Not better for:** native meeting notes, broad file automation, or a persistent desktop agent. ## Which alternative fits your Mac? Use this shorter decision tree: - **One app for chat, dictation, meetings, and agent tasks:** Clovy. - **Frontier cloud reasoning and coding:** Claude Desktop. - **Google Workspace and quick visual context:** Gemini. - **The hardest local boundary with a polished UI:** LM Studio. - **Open-source local chat:** Jan. - **Local document workspaces:** AnythingLLM. - **Private browser assistance:** Brave Leo. If your employer already uses ChatGPT Business or Enterprise, switching products may not improve governance. First compare the organization's retention, access, and audit requirements with what the alternative actually offers. ## Sources and verification This comparison was checked on July 20, 2026 using primary documentation: - [OpenAI macOS requirements](https://help.openai.com/en/articles/9395554-what-are-the-system-requirements-for-the-chatgpt-macos-app), [Work with Apps](https://help.openai.com/en/articles/10119604-work-with-apps-on-macos), and [Mac app retention](https://help.openai.com/en/articles/9268871-how-is-data-retained-in-the-macos-app) - [Claude Quick Entry](https://support.claude.com/en/articles/12626668-use-quick-entry-with-claude-desktop-on-mac), [Claude Desktop deployment](https://support.claude.com/en/articles/12611117-deploy-claude-desktop-for-macos), and [Claude Cowork](https://support.claude.com/en/articles/13345190-get-started-with-claude-cowork) - [Gemini desktop requirements](https://support.google.com/gemini/answer/17011627?hl=en) and [Gemini privacy controls](https://support.google.com/gemini/answer/13594961?hl=en) - [Google Workspace generative AI privacy](https://support.google.com/a/answer/14130944?hl=en) - [LM Studio offline operation](https://lmstudio.ai/docs/app/offline) - [Jan overview](https://www.jan.ai/docs) and [Mac requirements](https://www.jan.ai/docs/desktop/install/mac) - [AnythingLLM Desktop](https://anythingllm.com/desktop) - [Brave Leo](https://brave.com/leo/) and [Brave browser privacy](https://brave.com/privacy/browser/) - [Clovy FAQ](https://opensoftware.co/clovy/faq) and [Clovy API verification](https://opensoftware.co/verify) --- # Private alternatives to ChatGPT for work in 2026 > Six private ChatGPT alternatives compared by retention, training, local storage, model quality, and fit for confidential professional work. - Canonical URL: https://www.opensoftware.co/blog/private-alternatives-to-chatgpt-for-work - Published: Jul 20, 2026 - Updated: Jul 20, 2026 - Category: Guides - Author: OpenSoftware - Reading time: 12 min read - Topics: private ChatGPT alternative, private AI for work, ChatGPT alternative for business, confidential AI assistant, local AI assistant, zero retention AI ## Primary links - [Learn more about Clovy](https://www.opensoftware.co/clovy) - [Download Clovy](https://www.opensoftware.co/download/mac) - [Clovy community](https://t.me/+B4Z8KUqEsRc4ZGVh) - [Open source repository](https://github.com/open-software-network/os-clovy) - [Venice AI](https://venice.ai/) - [Hermes Agent](https://hermes-agent.nousresearch.com/) - [Hermes Agent source](https://github.com/NousResearch/hermes-agent) ChatGPT can be made more private than its defaults suggest. You can turn off model training, use Temporary Chat, or move work into ChatGPT Business. None of those choices makes the consumer app local, though. Files uploaded from the Mac app are stored in OpenAI's cloud, saved conversations follow OpenAI's retention rules, and Temporary Chats can be kept for up to 30 days for safety. That may be acceptable for everyday drafting. It may not be the boundary you want for contracts, customer notes, financial documents, source code, or recorded meetings. This guide compares six alternatives using a stricter question: **what happens to your work after you press send?** If you are looking specifically for native Mac integration, see the separate guide to [ChatGPT desktop alternatives for Mac](https://opensoftware.co/blog/chatgpt-desktop-alternatives-for-mac). If Claude is your current tool, use the [private Claude alternatives](https://opensoftware.co/blog/private-alternatives-to-claude) guide instead. ## How we evaluated the alternatives This is a documentation-based comparison, not a benchmark of model intelligence. We checked vendor documentation current on July 20, 2026 and evaluated: - Where saved chats, files, and recordings live. - Whether content can be used for training. - How long content may be retained. - Whether inference is local, zero retention, or ordinary cloud processing. - Whether administrators get retention, audit, and access controls. - Whether the product covers real work beyond a chat box. ## Quick answer | Alternative | Best for | Where history lives | Training default | Main limitation | |---|---|---|---|---| | [Clovy](https://opensoftware.co/clovy) | One private Mac workspace for chat, dictation, meetings, and agent work | On your Mac by default | Default Private models use zero retention and no training | macOS only; default inference still needs internet | | [Proton Lumo](https://proton.me/lumo) | Encrypted hosted chat | Device plus zero-access encrypted sync | No training | No native Mac app or desktop automation | | [Duck.ai](https://duck.ai) | Private access to several model families | Not stored by default; optional encrypted sync | No training under provider agreements | OpenAI and Anthropic may cache prompts in memory for up to one hour | | [Venice](https://venice.ai) | Choosing a privacy level per model | Local browser storage | Private mode is zero retention | Anonymous models may retain prompts upstream | | [LM Studio](https://lmstudio.ai) | Fully local chat and document work | On your Mac | Local models do not send prompts anywhere | Model quality and speed depend on your hardware | | [Jan](https://jan.ai) | Open-source local AI | On your Mac | Local use stays local | More setup than a hosted assistant | There is no universal winner. Hosted privacy services are easier. Local models create a harder data boundary. Clovy sits between them: the workspace and agent run on your Mac, while model calls use private hosted inference by default, with an optional local text-model endpoint for generation and agent work. ## What "private" should mean for work Privacy claims become confusing because vendors use different words for different controls. - **Training opt-out** means the provider says it will not use covered content to improve its models. It may still store the content. - **Zero retention** means the inference provider does not persist the covered prompt or response after processing, subject to documented transient caches and legal or abuse-prevention exceptions. - **Local history** means your saved conversations live on your device. It does not prove the model itself ran locally. - **Local inference** means the model runs on hardware you control and the prompt does not need to leave that hardware. - **Encrypted history** protects saved conversations. A hosted model still needs access to plaintext during inference unless the computation runs inside a verified encrypted environment. For confidential work, also ask about account records, metering, abuse monitoring, connected tools, web search, and deliberately submitted feedback. "No training" is not the same as "nothing is collected." ## Before switching: make ChatGPT safer If ChatGPT is already required by your team, start with its available controls. For individual accounts, turn off **Improve the model for everyone** before entering work content. The change applies to new conversations. Temporary Chat keeps a conversation out of history and model training, but OpenAI may retain a copy for up to 30 days for safety. Files uploaded through the macOS app are stored in the cloud rather than only on the Mac. ChatGPT Business and Enterprise do not train on business inputs and outputs by default. They are still cloud services with saved workspace data. Business chats remain until a user deletes them. Enterprise administrators can configure a workspace retention policy, subject to OpenAI's available limits. Those controls are meaningful. They solve a different problem from local storage or zero-retention inference. ## 1. Clovy: best private ChatGPT alternative for an integrated Mac workflow Clovy combines private chat, push-to-talk dictation into any app, meeting notes without a participant bot, and a local Hermes-based agent in one desktop workspace. Recordings, transcripts, notes, files, sessions, and agent memory are stored on your Mac by default. When Clovy needs hosted inference, its default Private models route through the open-source Clovy API to zero-retention Venice models. The covered provider does not retain the prompt or use it for training. Anonymous third-party models are optional and may apply different retention rules. The Clovy API runs inside an attested confidential virtual machine. The public [verification page](https://opensoftware.co/verify) lets you check the Clovy API deployment and source commit. That proof covers the Clovy API layer. It should not be read as proof of every upstream model provider, which is why the model's privacy label still matters. Clovy also supports an optional OpenAI-compatible local text model for generation and agent work. That can move text inference onto hardware you control. Dictation transcription and meeting-note generation are not fully local by default and still require a network connection. Zero retention describes prompt content on the default Private inference path. Clovy still keeps the account, billing, capacity, model-selection, request-timing, credit, and error records needed to operate the service. Material you deliberately send in an issue report is retained with that report. Connected tools, websites, and actions introduce their own recipient boundaries. The current Clovy repository is MIT-licensed and includes both the desktop app and Clovy API. The app supports macOS 14 or later on Apple Silicon and Intel. Capturing system audio for meetings requires macOS 14.2 or later; macOS 14.0 and 14.1 fall back to microphone-only recording. **Choose Clovy if:** you want one Mac workspace with local history, private hosted inference by default, and an agent that can work with local files and scheduled routines. **Do not choose Clovy if:** you need Windows today, require all speech processing to stay offline, or want a team knowledge base stored centrally in the cloud. **Pricing:** Free, Pro at $20/month, and Max at $100/month. Scheduled routines are available on Pro and above. Pricing changes usage and model access, not the privacy standard of the default Private tier. ## 2. Proton Lumo: best encrypted hosted chat Lumo is Proton's privacy-focused web and mobile assistant. It runs open models on servers controlled by Proton rather than sending prompts to third-party model providers. Proton says prompts are erased after processing, are not used for training, and do not appear in prompt or response logs. Saved history is protected with zero-access encryption, so Proton stores encrypted history but cannot decrypt it after it has been saved. The service still processes plaintext on Proton-controlled inference servers for the moment required to generate a response. Account, payment, security, and service records are separate from prompt-content logs. That architecture is strong for private hosted chat. It is not an offline model, a native Mac app, a system-wide dictation tool, or a desktop agent. **Choose Lumo if:** your primary need is private writing, summarization, and question answering in a browser. **Do not choose Lumo if:** you need a frontier model from a specific provider, Mac automation, meeting capture, or local file workflows beyond its supported integrations. ## 3. Duck.ai: best private access to multiple model families Duck.ai acts as a privacy layer between you and providers including OpenAI, Anthropic, Mistral, Azure, and Tinfoil. DuckDuckGo removes identifying metadata before sending a request and says its provider agreements prohibit training and require zero data retention. The details vary by model. OpenAI and Anthropic may hold prompt caches in memory for up to one hour, and Anthropic has limited legal and abuse-prevention exceptions. Tinfoil-hosted models add a Trusted Execution Environment that DuckDuckGo labels "zero provider visibility." Uploaded files and images are scanned for child sexual abuse material outside the encrypted Tinfoil path and may be retained and reported if flagged. Chats are not stored by default. Optional Sync and Backup stores them in end-to-end encrypted form. **Choose Duck.ai if:** you want to compare several model families without giving each provider your identity or direct account. **Do not choose Duck.ai if:** you need a native desktop workspace, offline use, persistent local agents, or one uniform guarantee across every available model. ## 4. Venice: best if you want to choose the privacy boundary per chat Venice stores conversation history in the browser and offers four privacy modes. - **Anonymous:** Venice hides your identity from a frontier provider, but the provider may store the prompt. - **Private:** the default for supported models, using Venice-controlled GPUs or zero-retention partners. - **TEE:** inference runs in a hardware-isolated environment with remote attestation. - **E2EE:** the prompt is encrypted on your device and decrypted only inside a verified TEE. E2EE trades away some features. It currently does not support web search, memory, file upload, or function calling. Venice also collects limited product and account event data even though it does not store prompt content on its servers. **Choose Venice if:** you want hosted private inference and understand the model-by-model privacy labels. **Do not choose Venice if:** you expect every frontier model in the catalog to have the same retention guarantee or need a native Mac desktop agent. ## 5. LM Studio: best polished local model workspace LM Studio can run downloaded models entirely offline. Local chats, document retrieval, and its OpenAI-compatible local server stay on the machine. Searching for and downloading models, checking for updates, web search, and optional cloud models require network access. The tradeoff is hardware. A smaller local model can be private and fast while still falling short of a large hosted model on difficult analysis, long documents, or tool use. You also own model selection, storage, updates, and backups. LM Studio now offers cloud models and web search under zero-retention terms. Selecting those features changes the system from local inference to hosted inference, even though the interface stays the same. **Choose LM Studio if:** your non-negotiable requirement is that ordinary chats and documents can remain on your Mac. **Do not choose LM Studio if:** you want a turnkey meeting-notes and dictation workflow or do not want to manage model files and memory constraints. ## 6. Jan: best open-source local ChatGPT-style app Jan is an Apache 2.0 licensed desktop application for macOS, Windows, and Linux. With a local model, it works offline, stores conversations and settings locally, and does not collect telemetry unless you allow it. Jan can also connect to cloud providers and external tools. Once you select a cloud model or networked integration, that provider's privacy terms apply. Open source does not make an external API local. On Mac, Jan currently requires macOS 13.6 or later and Apple Silicon. Model size determines the practical memory requirement. **Choose Jan if:** you want an auditable, no-account local assistant and are comfortable choosing open models. **Do not choose Jan if:** you need the quality of a specific frontier model without sending prompts to that provider, or want automatic meeting and dictation workflows. ## What about Claude, Gemini, Copilot, and Perplexity? These can be appropriate for organizational work, but their strongest privacy posture comes from plan terms and administration rather than local or zero-retention architecture. | Service | Stronger work configuration | Important limitation | |---|---|---| | Claude | Team or Enterprise content is not used for model training by default | Saved chats remain in Anthropic's cloud until deleted or an enterprise retention rule applies | | Gemini | Qualifying Google Workspace content is not used to train models outside Workspace without permission | Consumer Gemini has separate activity and retention controls | | Microsoft 365 Copilot | Enterprise Data Protection prevents prompts and responses from training foundation models | Prompts and responses are logged and governed through Microsoft Purview | | Perplexity | Enterprise data is not used for training and third-party providers use zero-retention terms | Consumer AI data retention is enabled by default unless you opt out | For regulated work, "not used for training" is only one control. Your organization may also need retention policies, audit logs, access controls, data residency, and a contract. A privacy-first consumer product is not automatically a compliance program. ## Requirements matrix for professional work | Requirement | Shortlist | What to verify before rollout | |---|---|---| | Contract review that must stay on one Mac | LM Studio or Jan with a local model | Model files, retrieval, tools, backups, and web search all remain local | | Confidential chat with hosted inference | Lumo, Duck.ai, Venice Private/TEE/E2EE, or Clovy Private models | Exact model badge, retention exceptions, abuse handling, and account metadata | | Meeting recordings and transcripts | Clovy | Recording consent, system-audio support, inference path, and local backup policy | | Centralized company governance | ChatGPT Enterprise, Claude Enterprise, Gemini for Workspace, or Microsoft 365 Copilot | Contract terms, admin retention, eDiscovery, residency, and incident response | | Local agent acting on files | Clovy, Jan, or a local-model stack | Approval model, sandbox boundary, connected tools, and whether the selected model is local | ## Decision guide Choose based on the boundary you actually need: 1. **Everything must remain on hardware you control:** use LM Studio or Jan with a local model and local tools. 2. **You want private hosted chat with encrypted history:** use Proton Lumo. 3. **You want several frontier providers behind a privacy proxy:** use Duck.ai, checking the label and exceptions for each model. 4. **You want hosted inference with selectable TEE or E2EE modes:** use Venice. 5. **You want chat, dictation, meetings, and local agent workflows in one Mac app:** use Clovy. 6. **Your employer requires centralized governance:** evaluate the business or enterprise plan of the provider your organization already manages. ## Sources and verification Product policies change. This article was checked on July 20, 2026 against the vendors' primary documentation: - [OpenAI data controls](https://help.openai.com/en/articles/7730893-chatgpt-data-controls-faq) and [Mac app retention](https://help.openai.com/en/articles/9268871-how-is-data-retained-in-the-macos-app) - [OpenAI business data commitments](https://openai.com/business-data/) - [OpenAI chat and file retention](https://help.openai.com/en/articles/8983778-chat-and-file-retention-policies-in-chatgpt) - [Proton Lumo privacy](https://proton.me/support/lumo-privacy) - [Duck.ai privacy details and provider exceptions](https://duckduckgo.com/duckduckgo-help-pages/duckai/ai-chat-privacy) - [Venice privacy modes](https://venice.ai/privacy) - [Venice TEE and E2EE feature limits](https://docs.venice.ai/guides/features/tee-e2ee-models) - [LM Studio offline operation](https://lmstudio.ai/docs/app/offline) and [app privacy](https://lmstudio.ai/app-privacy) - [Jan privacy](https://www.jan.ai/docs/desktop/privacy) and [Mac requirements](https://www.jan.ai/docs/desktop/install/mac) - [Clovy privacy and product FAQ](https://opensoftware.co/clovy/faq) and [Clovy API verification](https://opensoftware.co/verify) --- # The best private alternatives to Claude in 2026 > Six private Claude alternatives compared on training, retention, local models, encrypted history, Mac support, writing, coding, and agent workflows. - Canonical URL: https://www.opensoftware.co/blog/private-alternatives-to-claude - Published: Jul 20, 2026 - Updated: Jul 20, 2026 - Category: Guides - Author: OpenSoftware - Reading time: 10 min read - Topics: private Claude alternative, Claude alternative privacy, local Claude alternative, Claude desktop alternative, private Claude replacement, Claude alternative for work ## Primary links - [Learn more about Clovy](https://www.opensoftware.co/clovy) - [Download Clovy](https://www.opensoftware.co/download/mac) - [Clovy community](https://t.me/+B4Z8KUqEsRc4ZGVh) - [Open source repository](https://github.com/open-software-network/os-clovy) - [Venice AI](https://venice.ai/) - [Hermes Agent](https://hermes-agent.nousresearch.com/) - [Hermes Agent source](https://github.com/NousResearch/hermes-agent) Claude is a strong writing, coding, and analysis product. Its consumer privacy controls are also more nuanced than the claim that it simply "trains on everything." On Claude Free, Pro, and Max, Anthropic uses chats and coding sessions for model improvement only if you allow it, apart from safety review and content you deliberately submit as feedback. Saved chats remain in your account until you delete them. Deletion removes a chat from the product immediately and schedules backend deletion within 30 days. If you allow model improvement, eligible data may be held in de-identified training pipelines for up to five years. Turning Model Improvement off cannot undo training already in progress or models already trained. Anthropic may retain inputs and outputs flagged by automated safety systems for up to two years and the related trust and safety classification scores for up to seven years. Legal, abuse-prevention, and feedback exceptions can also apply. Claude for Work does not train on organizational content by default. It is still a retained cloud workspace unless your enterprise retention policy says otherwise. So the case for an alternative is not that Claude has no privacy controls. It is that you may want a different architecture: local inference, local history, zero retention, encrypted storage, or a desktop agent whose workspace lives on your Mac. This guide compares alternatives for the jobs Claude is commonly hired to do: long-document analysis, writing, coding, Cowork-style agent tasks, and work that must remain on a Mac. For company-wide governance, use [private ChatGPT alternatives for work](https://opensoftware.co/blog/private-alternatives-to-chatgpt-for-work). For native Mac ergonomics, see [ChatGPT desktop alternatives for Mac](https://opensoftware.co/blog/chatgpt-desktop-alternatives-for-mac). ## How we evaluated Claude replacements This is a documentation-based workflow comparison, not a claim that the models produce identical answers. We checked vendor sources current on July 20, 2026 and evaluated long-document support, writing and coding fit, file-based agent work, inference location, history storage, and retention. ## Quick answer | Alternative | Best for | Privacy model | Closest Claude-like strength | |---|---|---|---| | [Clovy](https://opensoftware.co/clovy) | Private work across chat, files, dictation, and meetings on Mac | Local-first workspace plus zero-retention hosted inference by default | Agentic work with local context | | [Proton Lumo](https://proton.me/lumo) | Private hosted writing and summarization | No prompt/response logs; no training; zero-access encrypted history | Simple web chat | | [Duck.ai](https://duck.ai) | Accessing Claude and other models through a privacy proxy | Anonymization plus provider agreements; OpenAI/Anthropic caches may remain in RAM for one hour | Choice of frontier models | | [Venice](https://venice.ai) | Choosing TEE or E2EE hosted inference | Four model-specific privacy modes | Hosted model variety | | [LM Studio](https://lmstudio.ai) | Fully local analysis and document chat | Local inference and local storage | Private long-document work, hardware permitting | | [Jan](https://jan.ai) | Open-source local chat and tools | Local models, local history, opt-in telemetry | Open, extensible desktop experience | No product exactly reproduces Claude's model behavior. If Claude's specific reasoning or writing quality is essential, the practical choice may be Claude with Model Improvement off, Claude Incognito chats, or Claude for Work. Incognito is not zero retention: Anthropic retains Incognito chats for 30 days by default, and an organization's retention policy can keep them longer. Team and Enterprise Incognito chats may also be available to administrators and exports. Choose an alternative when the privacy boundary matters more than using a specific model. ## What Claude's privacy settings do and do not do Turning Model Improvement off prevents eligible new and existing chats from being used for future training runs, according to Anthropic's current consumer documentation, except work already in an active training run or a completed model. Incognito chats are not used for model improvement even if the setting is enabled, but their default retention is 30 days. These controls do not make Claude local. Saved consumer chats and files still support the cloud product and remain until deleted. Claude Desktop Quick Entry, voice dictation, extensions, and Cowork also send work through Anthropic's service. Cowork cloud sessions run on Anthropic's servers, even when the desktop app provides access to a local file or browser. For organizations, Team and Enterprise provide no-training terms and administrative controls. Enterprise can apply custom retention, but the default is still a cloud workspace rather than zero-retention inference. ## 1. Clovy: best private Claude alternative for work on a Mac Clovy combines private chat with dictation, meeting notes, and a local Hermes-based agent. The workspace history is stored on your Mac by default, including recordings, transcripts, notes, files, sessions, and memory. The default model path is hosted, not fully offline. Clovy sends the selected content required for inference through the open-source Clovy API to Venice Private models with zero-retention and no-training terms. Optional Anonymous providers can have different retention rules. The model picker makes that distinction visible. The Clovy API runs in an attested confidential virtual machine. You can inspect the public repository and validate deployment evidence at [opensoftware.co/verify](https://opensoftware.co/verify). This verifies the Clovy API layer. It does not turn every upstream model into an attested service. Clovy can also connect to an optional OpenAI-compatible local text model for generation and agent work. That gives power users a local inference path without claiming that Clovy's dictation or meeting transcription is fully offline. The zero-retention claim covers prompt content on the default Private inference path. Clovy still stores account, billing, capacity, model-selection, request-timing, credit, and error records. Issue reports keep what the user deliberately submits. Connected tools and web actions have their own recipient boundaries. Compared with Claude Desktop, Clovy is less about one frontier model and more about keeping the working context on the Mac. Its agent can work with local files, ask for approval before risky actions, and, on Pro or Max, run scheduled routines. The public codebase is MIT-licensed. Clovy supports macOS 14 or later on Apple Silicon and Intel. System-audio meeting capture requires macOS 14.2 or later. **Choose Clovy if:** you want a private Mac workspace spanning voice, meetings, documents, and agent tasks. **Keep Claude if:** Claude's exact model quality, Cowork ecosystem, or an organization-managed Claude workspace is the priority. ## 2. Proton Lumo: best simple hosted alternative Lumo runs open models on Proton-controlled servers. Proton says it does not keep prompt or response logs, use chats for training, or send requests to third-party model providers. After inference, prompt content is erased from the processing system. Account, payment, security, and service records are separate. Saved history uses zero-access encryption. Proton's server temporarily decrypts the active conversation for inference, then saves history in a form Proton says it cannot decrypt. Lumo is a focused chat product available on the web and mobile. It does not currently replace Claude Desktop, Claude Code, or Cowork. Its value is a clearer hosted privacy model for drafting, summarization, and ordinary questions. **Choose Lumo if:** you want encrypted hosted chat without managing local models. **Keep Claude if:** you need Claude's coding tools, model quality, integrations, or agentic work surfaces. ## 3. Duck.ai: best way to use Claude models behind a privacy proxy Duck.ai is unusual because it can provide access to selected Anthropic models without sending Anthropic your IP address or direct account identity. DuckDuckGo says it strips identifying metadata and contracts with providers for no training and zero data retention. Anthropic prompt caching may keep chat content in memory for up to one hour, and limited legal or misuse exceptions apply. Uploaded images and files are scanned for child sexual abuse material outside the Tinfoil encrypted path and may be retained and reported if flagged. That is stronger than an ordinary consumer Claude session, but it is not local inference. Duck.ai also offers OpenAI, Mistral, Azure, and Tinfoil-hosted models. Tinfoil models add a Trusted Execution Environment and "zero provider visibility." The privacy guarantee therefore depends on the selected model, not merely the Duck.ai brand. **Choose Duck.ai if:** you want private access to multiple frontier model families or specifically want Claude behind an anonymizing proxy. **Keep Claude if:** you depend on Projects, Cowork, Claude Code, long-lived account context, or Anthropic's newest features on release day. ## 4. Venice: best configurable hosted privacy Venice offers Anonymous, Private, TEE, and E2EE model modes. Anonymous mode hides your identity from providers such as Anthropic but explicitly tells you to assume the upstream provider stores the prompt. It is not the mode to choose when retention is the reason you are leaving Claude. Private mode uses Venice-controlled GPUs or zero-retention partners. TEE mode adds hardware-isolated inference with attestation. E2EE encrypts the prompt on the device and decrypts it only inside a verified TEE, but gives up features such as web search and memory. Conversation history stays in local browser storage. Venice still records limited account and product event data, which is separate from prompt content. **Choose Venice if:** you want to trade model selection and features for stronger, clearly labeled hosted privacy modes. **Keep Claude if:** you need Anthropic's complete feature set and are satisfied with its account and retention controls. ## 5. LM Studio: best local alternative for documents and private analysis LM Studio runs downloaded open models entirely on your Mac. With local models selected, prompts, documents, retrieval, and chat history stay on the machine. It can also expose an OpenAI-compatible local server to other applications. The privacy boundary is strong because there is no hosted inference request. The capability tradeoff can also be large. Claude's strongest models run on data-center hardware. A local alternative must fit in the memory and compute available on your Mac. For routine drafting and private document search, a good smaller model may be enough. For difficult reasoning, very long context, or advanced tool use, it may not match Claude. Optional LM Studio cloud models and web search use zero-retention processing, but they are no longer local. **Choose LM Studio if:** local inference is the requirement and you are willing to test models against your real workload. **Keep Claude if:** predictable frontier performance matters more than keeping every prompt on-device. ## 6. Jan: best open-source local alternative Jan is an open-source desktop platform for local models and tools. It works offline after model download, keeps app data in a local folder, requires no account for local use, and makes analytics optional. Jan can connect to Claude, GPT, and other cloud APIs. Those requests follow the provider's terms. A local-first interface does not protect a prompt sent to a remote model. Jan currently supports Apple Silicon Macs on macOS 13.6 or later. It also runs on Windows and Linux. **Choose Jan if:** you want an auditable local assistant, open models, and an extensible tool system. **Keep Claude if:** you prefer a managed service and do not want to select, download, and maintain models. ## Which alternative replaces the part of Claude you use? | Claude job | Strongest private alternative | Tradeoff | |---|---|---| | Reviewing a contract that cannot leave one Mac | LM Studio or Jan with a local model | Local models may be weaker on long context and difficult reasoning | | Drafting and summarizing in a hosted chat | Proton Lumo | Simpler product and different open models | | Using Claude or other frontier models without a direct provider identity | Duck.ai | Browser-first, no Projects, Cowork, or Claude Code | | Hardware-verified hosted inference | Venice TEE or E2EE | E2EE gives up web search, memory, file upload, and function calling | | Working across local files, meetings, voice, and scheduled tasks | Clovy | macOS only; speech workflows are not fully offline by default | | Claude Code or Cowork with organizational governance | Claude for Work | Retained Anthropic cloud workspace rather than local or zero retention | Do not use a consumer privacy comparison as a substitute for your employer's security review. Legal privilege, regulated records, data residency, contractual confidentiality, and incident response require more than a no-training toggle. ## Sources and verification This article was checked on July 20, 2026 against primary vendor documentation: - [Anthropic consumer training policy](https://privacy.anthropic.com/en/articles/10023580-is-my-data-used-for-model-training), [consumer retention](https://privacy.claude.com/en/articles/10023548-how-long-do-you-store-my-data), and [Claude for Work data use](https://privacy.anthropic.com/en/articles/7996885-how-do-you-use-personal-data-in-model-training) - [Claude Incognito retention and organization visibility](https://support.claude.com/en/articles/12260368-use-incognito-chats) - [Claude Quick Entry](https://support.claude.com/en/articles/12626668-use-quick-entry-with-claude-desktop-on-mac) and [Claude Cowork](https://support.claude.com/en/articles/13345190-get-started-with-claude-cowork) - [Claude Cowork architecture](https://support.claude.com/en/articles/14479288-claude-cowork-architecture-overview) - [Proton Lumo privacy](https://proton.me/support/lumo-privacy) - [Duck.ai privacy details](https://duckduckgo.com/duckduckgo-help-pages/duckai/ai-chat-privacy) - [Venice privacy modes](https://venice.ai/privacy) - [LM Studio offline operation](https://lmstudio.ai/docs/app/offline) - [Jan overview](https://www.jan.ai/docs) and [privacy approach](https://www.jan.ai/docs/desktop/privacy) - [Clovy FAQ](https://opensoftware.co/clovy/faq) and [Clovy API verification](https://opensoftware.co/verify) --- # The best AI apps for Mac in 2026 > Seven Mac AI apps ranked by use case with honest tradeoffs on privacy, pricing, and limits: Clovy, ChatGPT, Claude, Wispr Flow, Superwhisper, Granola, and Otter. - Canonical URL: https://www.opensoftware.co/blog/best-ai-apps-for-mac - Published: Jul 16, 2026 - Updated: Jul 16, 2026 - Category: Guides - Author: OpenSoftware - Reading time: 13 min read - Topics: best AI apps Mac 2026, Mac AI assistant, private AI Mac, AI dictation Mac, AI meeting notes, Clovy, ChatGPT desktop, Claude ## Primary links - [Learn more about Clovy](https://www.opensoftware.co/clovy) - [Download Clovy](https://www.opensoftware.co/download/mac) - [Clovy community](https://t.me/+B4Z8KUqEsRc4ZGVh) - [Open source repository](https://github.com/open-software-network/os-clovy) - [Venice AI](https://venice.ai/) - [Hermes Agent](https://hermes-agent.nousresearch.com/) - [Hermes Agent source](https://github.com/NousResearch/hermes-agent) Most "best AI apps" lists treat every app as roughly equal and sort them by star rating. That's not useful anymore. In 2026, what separates a genuinely good Mac AI app from a risky one is *what it does with your data* before you ever see an output. Your voice, your meetings, your documents: these go somewhere when you hit record or send a prompt. Sometimes it's a zero-retention model. Sometimes it's a training pipeline. Often, you can't tell from the marketing page. This guide cuts through that. It covers the seven apps that matter most (Clovy, ChatGPT desktop, Claude, Wispr Flow, Superwhisper, Granola, and Otter), ranked by use case, with honest tradeoffs on privacy, pricing, and limits. ## Quick answer: best AI apps for Mac in 2026 by use case | Use case | Best privacy-first pick | Cloud/hybrid alternative | |---|---|---| | System-wide AI dictation | Clovy | Wispr Flow, Superwhisper | | Bot-free meeting notes | Clovy | Granola | | General chat and research | Clovy, Claude | ChatGPT desktop | | Local file and workflow agent | Clovy | -- | **Privacy level rubric:** - **Local-first:** files, sessions, and memory stay on your device by default, with model calls routed through privacy-preserving infrastructure - **Cloud/hybrid:** some or all processing routes through vendor servers; privacy controls may exist but require configuration - **Unclear:** retention and training defaults are not well documented on the product page Clovy sits in the local-first row across all four use cases. It's the only app on this list that bundles dictation, meeting notes, chat, and a local agent in a single private workspace. ## How to pick the right Mac AI app (what actually matters) Before you look at any feature list, ask these four questions: 1. **What gets stored and for how long?** Some apps delete raw audio immediately but keep transcripts indefinitely. ChatGPT Record mode, for example, deletes source audio from OpenAI's servers right after transcription (per the OpenAI Help Center), but the generated transcript and any saved canvases are retained under your workspace's settings. That's a meaningful distinction. 2. **Does the app train on your content?** "Opt-out" and "zero retention" are not the same thing. Opt-out means training is on by default and you have to disable it. Zero retention means the covered provider does not retain the request or use it for training after inference. 3. **Where does processing happen?** On-device Whisper models versus a cloud API endpoint have very different risk profiles, especially for legal, medical, or financial content. 4. **What are the real limits?** Word caps, meeting-minute limits, and note history windows matter. Wispr Flow's free desktop tier caps transcription at 2,000 words per week. Granola's free plan restricts note history to a rolling 30 day window. These constraints affect whether the app is actually usable for daily work. ## 1. Clovy by OpenSoftware (best overall for privacy-conscious Mac users) [Clovy](https://opensoftware.co/clovy) is a private AI assistant for the Mac built by OpenSoftware. It bundles four things that most people run as separate apps: AI chat, [system-wide push-to-talk voice dictation](https://opensoftware.co/clovy/private-dictation-for-mac), [automated meeting notes without a bot joining your call](https://opensoftware.co/clovy/ai-meeting-notes-without-a-bot), and a local agent for file analysis and workflow automation. Files, sessions, memory, and agent state are stored on your device by default. When prompts do leave the device, they route to zero-retention Venice models by default on the Private tier, meaning the provider does not retain the request or use it for training after inference. The Clovy API is open source, MIT licensed, and runs inside a Trusted Execution Environment (TEE) with cryptographic attestation you can verify yourself at [opensoftware.co/verify](https://opensoftware.co/verify). That last part is unusual. Most apps ask you to trust their privacy policy. Clovy lets you verify it cryptographically. The local agent is built on the open-source Hermes framework and can work with PDFs, contracts, receipts, and other documents. The files themselves stay on your device; model calls route through the Clovy API under the privacy tier you select, with zero-retention Venice models as the default. If you want direct control over model terms, Clovy also supports bringing your own Venice API key. **Pricing:** Free (Hobby tier, for real work with light usage); Pro at $20/month. Requires macOS 14 or later on Apple Silicon or Intel. **Best for:** Knowledge workers, consultants, and anyone handling sensitive documents who want verifiable privacy across chat, dictation, and meetings in one place. ## 2. Superwhisper (best dedicated dictation app, offline-capable) Superwhisper is a focused AI dictation app for Mac. Its headline feature is offline, on-device transcription using local Whisper models, making it a solid choice for voice to text for Mac without touching the cloud at all. The fully on-device configuration (local speech-to-text plus local language models) requires the Pro plan and runs best on Apple Silicon. For Superwhisper vs Wispr Flow comparisons, Superwhisper's key advantage is the offline mode. On privacy specifics: cloud model usage on the Pro tier is covered by named zero data retention agreements with Anthropic, OpenAI, Deepgram, and Groq (per Superwhisper's docs, ZDR applies to API usage only), while recordings accumulate on your disk with no auto-delete. Worth noting: its privacy policy was last updated in June 2024, before the cloud model expansion. Superwhisper also includes a manual meeting mode: you start a recording yourself and it produces a summary, with speaker separation on Pro. There's no automatic meeting detection and no bot. **Pricing:** Free tier available; Pro at $8.49/month, $84.99/year, or $249.99 lifetime (as of July 2026). **Best for:** Dictation-first workflows where on-device offline processing is a priority, with basic manual meeting capture as a bonus. No agent. For a detailed side-by-side, see [Clovy vs Superwhisper](https://opensoftware.co/clovy/vs/superwhisper). ## 3. Wispr Flow (best dictation app for multi-app integration) Wispr Flow positions itself as a voice layer for any Mac app, with broad integration across browsers, editors, and messaging tools. Transcription always happens in the cloud; there is no offline mode. The defaults matter here: Privacy Mode is off and Cloud Sync is on out of the box, which means audio and transcripts are stored on US servers and may be used to train Wispr's models until you change the settings. Zero data retention requires both Privacy Mode turned on and Cloud Sync turned off, and that combination is the default only on Enterprise and HIPAA plans (per Wispr Flow's data controls page, June 2026). The free desktop tier caps transcription at 2,000 words per week, which is limiting for daily use. **Pricing:** Free tier with word limits; Pro at $15/month ($12/month annual). **Best for:** Users who want seamless dictation across many apps and don't need meeting notes or an agent. Before using it with sensitive content, turn Privacy Mode on and Cloud Sync off. For a direct comparison on privacy defaults, see [Clovy vs Wispr Flow](https://opensoftware.co/clovy/vs/wispr-flow). ## 4. Granola (best bot-free meeting notes, cloud-stored) Granola calls itself an "AI notepad for meetings" and it captures meetings without sending a bot into your call, which is a real differentiator versus Otter's OtterPilot default. It listens to system audio and mic, then generates structured notes afterward. Audio is discarded after transcription, which is a genuine privacy strength. The rest of the privacy picture is mixed. Granola stores transcripts and notes on US cloud servers, and it trains its own models on de-identified data by default (you can opt out in settings; it's disabled by default for Enterprise). The free plan limits note history to a rolling 30 day window. Business pricing runs $14/user/month (as of July 2026). For a Granola vs Otter decision, both hold SOC 2 Type II, so that's a wash. Granola wins on the no-bot default; Otter's genuine strength is live transcription with named speaker identification, which Granola doesn't do. Neither keeps your notes on your local device by default. **Pricing:** Free (rolling 30 day history window); Business at $14/user/month. **Best for:** Teams that want clean meeting notes without a bot participant and don't have strict data residency requirements. For a full breakdown, see [Clovy vs Granola](https://opensoftware.co/clovy/vs/granola). ## 5. Otter.ai (best for live transcription and CRM-connected meeting workflows) Otter brings serious enterprise credentials: SOC 2 Type II, a HIPAA BAA on Enterprise plans, and integration guides for Salesforce and HubSpot in its help center. Its genuine standout feature is live transcription with named speaker identification. Capture works two ways. The default most people know is OtterPilot, a bot that joins your Zoom, Meet, or Teams call via calendar integration, visible to everyone in the meeting. Otter also offers bot-free desktop capture from system audio, which is worth crediting. The [Clovy vs Otter comparison](https://opensoftware.co/clovy/vs/otter) goes deeper on the tradeoffs. On training: Otter's trust page says de-identified recordings and transcripts are used to train its models automatically, with no consumer opt-out shown on that page. That's the fact to weigh, not the badges. **Pricing:** Free (300 minutes/month, 30 minutes per conversation); Pro at $16.99/month or $8.33/month annual; Business at $30/month or $19.99/month annual. **Best for:** Teams that need live transcription, named speakers, CRM sync, and shared meeting libraries, and are comfortable with cloud storage and default training. ## 6. ChatGPT desktop (best general chat, weaker on Mac-native privacy) ChatGPT desktop is an obvious choice for general research, drafting, and Q&A. Its Record mode is worth discussing specifically because it illustrates the audio vs transcript retention distinction clearly. According to the OpenAI Help Center, when you use ChatGPT Record mode, the original source audio is deleted from OpenAI's servers right after transcription. But the resulting transcript and structured summaries (saved as "canvases") are stored under your workspace's retention settings, which depend on your account type and any active data controls. So the audio is gone quickly; the content derived from it may not be. Record mode is macOS-only, requires a paid plan (Plus and up), and caps sessions at 240 minutes. For consumer accounts, training is enabled by default: the "improve the model for everyone" setting is on unless you turn it off, and opting out is not retroactive. ChatGPT desktop doesn't do system-wide dictation, and its bot-free meeting capture comes only through paid Record mode, so it's a narrower surface on the Mac compared to an integrated workspace like Clovy. The Mac app also requires Apple Silicon; there's no Intel support. **Pricing:** Free tier; Plus at $20/month. **Best for:** General-purpose chat and writing assistance. Less suitable for sensitive professional workflows without careful data control configuration. Compare the two directly with [Clovy vs ChatGPT desktop](https://opensoftware.co/clovy/vs/chatgpt-desktop). ## 7. Claude desktop (best for long-context research and drafting) Claude (Anthropic) is strong for contract review, document summarization, and multi-step research, helped by a long context window (Anthropic's platform documentation describes context windows of up to 1 million tokens on recent models, enough to hold lengthy contracts or dozens of papers in one request). The Mac app is clean and fast. On privacy: since Anthropic's consumer terms update, chats on consumer plans (Free, Pro, and Max) are used for model training by default unless you turn the setting off in your privacy settings, per Anthropic's own policy pages. Claude for Work is covered by separate commercial terms. Like ChatGPT, Claude desktop doesn't offer system-wide dictation or meeting capture, so it's a chat-only surface. **Pricing:** Free tier; Pro at $20/month; Claude for Work varies. **Best for:** Long-context writing, research, and analysis tasks. Pair with a dictation tool if you want voice input. ## Side-by-side comparison | | Clovy | Superwhisper | Wispr Flow | Granola | Otter | ChatGPT desktop | |---|---|---|---|---|---|---| | System-wide dictation | Yes | Yes | Yes | No | No | No | | Meeting notes | Yes | Manual meeting mode | No | Yes | Yes | Record mode (paid, 240 min cap) | | Bot-free capture | Yes | Yes (manual) | N/A | Yes | Yes (desktop capture; OtterPilot bot also available) | Yes (Record mode) | | Local-first storage | Yes (default) | Partial (offline mode, Pro) | No | No | No | No | | Zero-retention default | Yes (Private tier, Venice models) | Local mode sends nothing to servers; cloud API use has ZDR agreements | No (Privacy Mode on + Cloud Sync off required, not default) | No (de-identified training on by default) | No (automatic de-identified training) | No (consumer training on by default, opt out available) | | Local file agent | Yes | No | No | No | No | No | | Free tier | Yes | Yes | Yes (2k words/wk) | Yes (30 day history) | Yes (300 min/mo) | Yes | | Paid pricing | $20/month | $8.49/month | $15/month | $14/user/month | $16.99/month | $20/month | | macOS requirement | macOS 14+ (Apple Silicon or Intel) | macOS 13.3+ | macOS varies | macOS varies | macOS varies | macOS 14+ (Apple Silicon only) | ## FAQ: best AI apps for Mac 2026 **Does the app train on my dictation or meeting notes?** It depends entirely on the app and your plan tier. ChatGPT's consumer training is enabled by default; you can opt out, but the opt-out isn't retroactive. Clovy routes to zero-retention Venice models by default on the Private tier, so nothing is retained for training there; opt-in Anonymous-tier providers may retain what they receive. Wispr Flow's zero-retention state requires Privacy Mode on and Cloud Sync off, neither of which is the default outside Enterprise/HIPAA plans. Always check the "data controls" or "privacy" settings immediately after installing any AI app. **Is audio deleted after transcription?** Sometimes. ChatGPT Record deletes source audio right after transcription but keeps the transcript. Granola discards audio but stores notes in the cloud. Clovy keeps transcripts on your device. "Audio deleted" doesn't mean your content is private if the transcript goes to a cloud training pipeline. **Are there word or minute limits on free plans?** Yes. Wispr Flow's free desktop tier caps at 2,000 words per week. Granola's free plan limits note history to a rolling 30 days. Otter's free plan caps at 300 minutes per month. Clovy's free Hobby tier is built for real work with light usage, and it includes dictation and meeting notes with the same privacy standard as Pro. **Do these apps send my audio to servers?** Superwhisper's offline mode processes audio entirely on-device using local Whisper models (a Pro feature). Clovy stores your workspace history on your device and routes model calls, including audio for transcription, through the Clovy API under your selected privacy tier, with zero-retention Venice models as the default. Most other apps in this list send audio or transcripts to their servers to some extent. **What hardware do I need for local-first options?** Superwhisper's on-device mode runs well on Apple Silicon. Clovy requires macOS 14+ on Apple Silicon or Intel. For running large local models via LM Studio or Ollama alongside a chat frontend, Apple Silicon with 16GB+ unified memory is practical. **Is macOS 14+ actually required for Clovy?** Yes. Clovy requires macOS 14 (Sonoma) or later, for both Apple Silicon and Intel Macs. If you're on an older macOS, Superwhisper or Wispr Flow may still run, but Clovy won't. **Can I bring my own API key?** Clovy supports bringing your own Venice API key in Settings, giving you a direct relationship with the model provider handling your requests. ## The bottom line The best AI apps for Mac in 2026 aren't the ones with the most features. They're the ones that are honest about where your data goes and give you a default you can actually trust. For dictation only: Superwhisper (offline, on Pro) or Wispr Flow (with Privacy Mode on and Cloud Sync off) are solid single-purpose picks. For meeting notes without a bot: Granola works well if cloud storage is acceptable. For general chat: Claude is strong on long-context work; ChatGPT is familiar and capable. If you want all four surfaces (chat, dictation, meeting notes, local agent) in one place, and you're comfortable with hosted private inference you can verify, Clovy is a strong fit. The free Hobby tier covers real work with light usage, the Pro tier is $20/month, and the TEE-backed cryptographic attestation at [opensoftware.co/verify](https://opensoftware.co/verify) means you don't have to take the privacy claims on faith. --- # The best AI productivity tools for macOS > The best Mac AI tools for meeting notes, dictation, and desktop automation in 2026, organized by workflow with privacy defaults called out. - Canonical URL: https://www.opensoftware.co/blog/best-ai-productivity-tools-for-macos - Published: Jul 16, 2026 - Updated: Jul 16, 2026 - Category: Guides - Author: OpenSoftware - Reading time: 10 min read - Topics: AI productivity tools Mac, AI meeting notes Mac, AI dictation, Mac AI tools 2026, Clovy, Granola, Otter, Wispr Flow, Raycast ## Primary links - [Learn more about Clovy](https://www.opensoftware.co/clovy) - [Download Clovy](https://www.opensoftware.co/download/mac) - [Clovy community](https://t.me/+B4Z8KUqEsRc4ZGVh) - [Open source repository](https://github.com/open-software-network/os-clovy) - [Venice AI](https://venice.ai/) - [Hermes Agent](https://hermes-agent.nousresearch.com/) - [Hermes Agent source](https://github.com/NousResearch/hermes-agent) If your day runs on back-to-back calls, half your mental energy goes to remembering what was decided and who said they'd handle it. The right AI productivity tools on macOS can handle that for you, but picking the wrong one means audio you didn't intend to share, bots awkwardly showing up as meeting participants, or a $30/month bill for features you only half use. This guide cuts through the noise. The tools below are organized by the three workflows where AI actually earns its keep on a Mac: meeting notes, voice dictation, and desktop automation. ## Quick answer: the shortlist for 2026 | Tool | Category | macOS native | Privacy stance | Starting price | |---|---|---|---|---| | **Clovy (OpenSoftware)** | Meeting notes + dictation + agent | Yes (14+) | Local-first workspace history; zero-retention Venice routing by default; TEE-verified Clovy API | Free / $20/mo | | **Granola** | Meeting notes | Yes | Cloud storage (US AWS); audio discarded after transcription | Free / $14/user/mo | | **Otter.ai** | Meeting notes | Yes (Mac and Windows desktop apps) | US AWS cloud; trains on de-identified data by default; admin-set retention available | Free tier available | | **Wispr Flow** | Dictation | Yes | Cloud-only transcription; zero retention with Privacy Mode on and Cloud Sync off | Free (2,000 words/wk) / $15/mo | | **Superwhisper** | Dictation | Yes | Fully on-device option (Pro); ZDR agreements for cloud API usage | Free / $8.49/mo or $249.99 lifetime | | **Raycast** | Launcher with AI | Yes | Cloud AI via hosted models | Free / Pro from $8/mo | | **ChatGPT desktop** | Chat + Record | Yes (Apple Silicon only) | Audio deleted post-transcription; transcripts stored on OpenAI servers | Free / Plus | ## How to pick without regretting it Before downloading anything, ask four questions about every tool you evaluate: 1. **Where does your audio go?** Is it processed on-device, on a private server, or a shared cloud? 2. **Is retention configurable?** Can you set how long transcripts are stored, or delete them on demand? 3. **Does it train on your data?** Defaults vary by product: Otter trains on de-identified recordings and transcripts automatically, Granola trains on de-identified data by default with an opt-out in settings, and Wispr Flow may use stored audio and transcripts for training until you change the defaults. Check the settings, not just the marketing page. 4. **Can you verify the privacy claims?** Badges (SOC 2, HIPAA) are a starting point, not an endpoint. One more distinction worth knowing: **bot-based vs. bot-free** meeting capture. Bot-based tools join as a named participant, which everyone can see in the participant list, so disclosure is harder to miss. Bot-free tools capture audio directly from your Mac, so nothing joins the call. Either way, recording-consent rules vary by jurisdiction, so check the rules that apply to your calls. The bot-free approach matters most on client calls and confidential interviews where a stray participant looks unprofessional. ## Category 1: AI meeting notes on Mac ### 1. Clovy by OpenSoftware, best for privacy-conscious knowledge workers with back-to-back meetings Clovy captures [AI meeting notes without a bot](https://opensoftware.co/clovy/ai-meeting-notes-without-a-bot) joining the call. It records from your Mac's audio, generates a transcript, and surfaces action items, with your workspace history stored on your device. Prompts route to zero-retention Venice models by default on the Private tier (opt-in Anonymous-tier models may retain what they receive), and the open-source, MIT-licensed Clovy API runs in a Trusted Execution Environment (TEE). You can verify that claim yourself at [opensoftware.co/verify](https://opensoftware.co/verify). Clovy also bundles system-wide voice dictation and a local desktop agent, so it's not just a meeting tool; it's your entire private AI workspace. Free to start; Pro is $20/month on macOS 14+ (Intel and Apple Silicon). ### 2. Granola, best for teams who want an AI notepad without a bot Granola positions itself as an AI notepad that captures meetings without a bot, and it discards audio after transcription. Notes and transcripts are stored in the cloud (US AWS), which suits teams sharing context; the Business plan adds Notion and Slack integrations. The free Basic tier offers limited meeting history; Business is $14/user/month with unlimited meeting notes and history (per granola.ai/pricing). It's a solid pick if cloud storage is acceptable and you want a clean, polished notes UI. See how it stacks up in a [Clovy vs Granola](https://opensoftware.co/clovy/vs/granola) breakdown. ### 3. Otter.ai, best for teams that need admin retention controls and live transcription Otter offers two capture modes: the OtterPilot bot can join Zoom, Meet, and Teams calls through calendar integrations, and there's also bot-free desktop capture from system audio on its Mac and Windows apps. Live transcription with named speaker identification is its genuine strength. Data is stored on US AWS cloud, and per Otter's trust page, de-identified recordings and transcripts train its models automatically, with no consumer opt-out shown there. Workspace admins can request a custom data retention period that automatically deletes conversations when it expires (per Otter's Help Center article "Set a custom Data Retention policy"; it's arranged through an Otter account manager). Read the [Clovy vs Otter.ai](https://opensoftware.co/clovy/vs/otter) comparison if you're deciding between the two. ## Category 2: AI dictation for Mac ### 4. Wispr Flow, best for professionals who want polished cloud dictation Wispr Flow's data controls page documents its architecture plainly: transcription always occurs in the cloud, and by default Privacy Mode is off and Cloud Sync is on, meaning audio and transcripts are stored on US servers and may be used to train Wispr's models until you change the settings. Zero data retention requires Privacy Mode on and Cloud Sync off, and that combination is the default only for Enterprise and HIPAA plans. On compliance, Wispr Flow completed a SOC 2 Type I re-audit in April 2026. It's built for continuous dictation into any app, which suits salespeople and lawyers who want to speak directly into their CRM or email client, and there's a free tier with 2,000 words per week on desktop. Compare defaults side by side in [Clovy vs Wispr Flow](https://opensoftware.co/clovy/vs/wispr-flow). ### 5. Superwhisper, best for offline voice-to-text on Mac Superwhisper offers a real fully on-device configuration: local speech-to-text plus local language models, both on the Pro plan, running on Apple Silicon. That's a strong option if you're frequently on planes or in low-connectivity environments, and it's the strongest local setup in the dictation category. Cloud models (also Pro) are covered by zero data retention agreements with the API providers. Pricing is flexible: a free tier, then Pro at $8.49/month, $84.99/year, or $249.99 for a lifetime license. It also has a manual meeting mode with recording and summaries (no meeting detection, no bot), though it lacks the desktop agent features of a full workspace tool. For pure offline dictation it's one of the clearest options. ## Category 3: AI agents and automation on the desktop ### 6. Raycast, best for Mac power users who want an AI launcher Raycast is a macOS launcher and command interface with AI features layered in: Quick AI for questions, AI chat, and custom AI commands for repetitive tasks, with access to hosted models from multiple providers on the Pro plan (from $8/month; a free tier covers the core launcher). If your automation needs are mostly quick lookups, snippets, and one-shot commands, it's a polished choice. For document-heavy or recurring multi-step agent workflows, evaluate whether a launcher-style tool fits before committing. ### 7. ChatGPT desktop, best for general AI chat and ad-hoc transcription The ChatGPT desktop app includes a Record mode that captures meetings from system audio without a bot. It's available on paid plans (Plus and up), caps recordings at 240 minutes, and per the OpenAI Help Center deletes raw audio after transcription; transcripts and summaries live in your ChatGPT history on OpenAI servers, subject to your plan's retention settings. The Mac app is Apple Silicon only, which leaves Intel Mac users out. Worth having for general chat and research, but Record mode is an add-on to a chat app rather than a dedicated meeting workflow. See a detailed [Clovy vs ChatGPT desktop](https://opensoftware.co/clovy/vs/chatgpt-desktop) breakdown for what's different. ## Where Clovy fits if you want one app for all three workflows If managing three separate tools (one for meeting notes, one for dictation, one for automation) sounds like overhead, Clovy is the alternative. It covers all three categories in a single [local AI agent for Mac](https://opensoftware.co/clovy/local-ai-agent), and its dictation is covered in depth at [private dictation for Mac](https://opensoftware.co/clovy/private-dictation-for-mac): - **Push-to-talk dictation** into any app, system-wide, with transcripts stored on your device - **Bot-free meeting notes** captured from your Mac's audio, with summaries and action items - **A local desktop agent** built on the open-source Hermes framework, for file analysis, recurring reports, and workflow automation Privacy defaults are set conservatively out of the box. You don't have to hunt through settings to enable a "privacy mode." Power users can also bring their own Venice API key. All of that is explained on the [Clovy FAQ](https://opensoftware.co/clovy/faq). ## Pricing reality check Here's a cost comparison for someone running four meetings a day: - **Granola Business:** $14/user/month (meeting notes, cloud, no dictation or agent) - **Superwhisper Pro + Granola Business:** ~$22.49/month (dictation + meeting notes, still no agent) - **Clovy Pro:** $20/month (meeting notes + dictation + local agent + private chat, local-first with default zero-retention Venice routing through the Clovy API) Add a separate AI launcher subscription like Raycast Pro and the alternative stack climbs further. Clovy's $20/month Pro plan covers the full workflow in one subscription. ## FAQ **Are AI meeting note tools safe for confidential calls?** It depends on where audio is processed and how long it's retained. Bot-based tools send audio to cloud servers; bot-free tools like Clovy capture from your Mac's audio without anything joining the call, then route model calls under your selected privacy tier. Always check the retention policy and whether the tool trains on your data. **Do AI dictation tools keep audio after transcription?** It varies by product, so check each one. Wispr Flow stores audio and transcripts on US servers under its defaults; zero data retention requires Privacy Mode on and Cloud Sync off. Superwhisper's recordings accumulate on your disk (its fully local mode never sends audio to a server, and its cloud API usage is covered by ZDR agreements). Clovy keeps workspace history on your Mac and routes model calls through the Clovy API under the selected privacy tier. **Is ChatGPT Record audio deleted after transcription?** Yes, according to the OpenAI Help Center: audio from Record mode is used only for transcription and deleted afterward. Transcripts, however, are saved in chat history on OpenAI servers and subject to your plan's retention settings. **Which tools are bot-free on macOS?** Granola and Clovy capture audio from the Mac rather than joining as a bot participant, and Superwhisper's manual meeting mode records without a bot. Otter offers both: the OtterPilot bot joins via calendar integrations, and bot-free desktop capture from system audio is also available. **What should I check for HIPAA or SOC 2 readiness?** Don't rely only on badges. Look for a Business Associate Agreement (BAA) for HIPAA, a SOC 2 Type II audit report (not just the badge), configurable retention windows, and an explicit statement about training defaults. Otter has SOC 2 Type II and an Enterprise HIPAA BAA, Granola has SOC 2 Type 2, and Wispr Flow completed a SOC 2 Type I re-audit in April 2026. Clovy's privacy architecture is verifiable directly at opensoftware.co/verify. ## Where to start If you're not sure which category applies most to your work, the fastest path is to try Clovy free on a real project. The free Hobby tier covers all three workflows (chat, dictation, meeting notes), so you'll know within a week whether it fits. Upgrade to Pro at $20/month when you want more usage, extended agent sessions, all AI models, skills and deeper workflows, scheduled routines, and priority support. If you specifically need enterprise compliance controls or a team-sharing workflow, Otter.ai and Granola both have team tiers worth evaluating alongside Clovy's own roadmap. --- # Best voice to text software for Mac in 2026 > A practical guide to Mac voice to text in 2026: live dictation, file transcription, and meeting notes, with honest privacy comparisons. - Canonical URL: https://www.opensoftware.co/blog/best-voice-to-text-software-for-mac - Published: Jul 16, 2026 - Updated: Jul 16, 2026 - Category: Guides - Author: OpenSoftware - Reading time: 14 min read - Topics: voice to text Mac, dictation app Mac, Mac speech to text, private dictation, Clovy, MacWhisper, Superwhisper, Wispr Flow, Apple Dictation ## Primary links - [Learn more about Clovy](https://www.opensoftware.co/clovy) - [Download Clovy](https://www.opensoftware.co/download/mac) - [Clovy community](https://t.me/+B4Z8KUqEsRc4ZGVh) - [Open source repository](https://github.com/open-software-network/os-clovy) - [Venice AI](https://venice.ai/) - [Hermes Agent](https://hermes-agent.nousresearch.com/) - [Hermes Agent source](https://github.com/NousResearch/hermes-agent) Published July 16, 2026 ## Best voice to text software for Mac in 2026 (quick pick) Not every voice-to-text tool does the same thing. Before comparing apps, it helps to know which category you actually need: - **Live dictation into any app:** Clovy. Hold the fn key, speak, and polished text lands at your cursor in any Mac app. Free to start; Pro is $20/month. - **Transcription of audio or video files:** MacWhisper. It transcribes existing recordings on your Mac using locally run models like Whisper and Parakeet, with optional cloud providers. Version 14 launched in July 2026 with a new transcript editor and faster performance. - **Meeting notes without a bot:** Clovy. It records from your Mac's audio so no participant bot joins the call, and transcripts and notes are stored on your device. - **Offline-first needs:** MacWhisper (local transcription of files) or Superwhisper's fully on-device Pro configuration on Apple Silicon. Clovy is partially offline: your dictation history and notes stay readable without internet, but model calls for dictation and cleanup require a connection. Clovy's category is live push-to-talk dictation with AI writing cleanup and a private workspace. It is not primarily a file transcription tool. --- ## Dictation vs transcription vs meeting notes: pick the right tool These three terms get used interchangeably in most listicles, but they describe different workflows. | Mode | What it does | When to use it | |---|---|---| | Live dictation | Types text at your cursor in real time as you speak | Writing emails, messages, documents, code comments | | File transcription | Converts an existing audio or video file into text | Recorded interviews, lectures, podcasts | | Meeting mode | Records a live call and generates structured notes | Calls you are in but not typing during | Clovy is built for live dictation into any app, with meeting notes as an adjacent feature that uses the same private infrastructure. If you need to batch-transcribe a folder of audio files, MacWhisper is the more direct fit. --- ## What makes Clovy different for 2026 dictation Several Mac voice-to-text apps exist in 2026, and a few do live dictation reasonably well. Clovy's differentiation comes down to three things. **Push-to-talk with polished output.** The default shortcut is the fn key: hold it, speak, release. Clovy transcribes what you said and then runs a cleanup pass that removes filler words, adds correct punctuation, and applies the writing style you selected. You can choose standard, casual lowercase, or formal. The result is writing that reads like you would have typed it, not a raw transcript. For longer stretches, a hands-free toggle keeps the session open without holding a key. **Privacy that is on by default.** Audio and prompts leave your Mac for inference, but they route to zero-retention Venice models by default. That means the model provider stores nothing and trains on nothing. Your dictation history is stored on your Mac, not on OpenSoftware servers. OpenSoftware retains account, login, and billing records, plus anything you deliberately submit in an in-app issue report. **Privacy you can verify, not just trust.** Clovy is open source and MIT-licensed. The Clovy API, which routes model calls, runs in a Trusted Execution Environment (TEE) and serves cryptographic attestation at [opensoftware.co/verify](https://opensoftware.co/verify). You can confirm what code is running on the servers yourself. That level of verifiability is uncommon among dictation tools, most of which rely on policy statements rather than technical proof. --- ## How live dictation works in Clovy The workflow is simple by design: 1. Install Clovy on macOS 14 or later (Apple Silicon or Intel). 2. Click into any app where you want to type: an email draft, a chat input, a code comment, a terminal. 3. Hold the fn key and speak naturally. 4. Release the key. Clovy processes the audio, cleans it up, and types the result at your cursor. Hands-free mode is available for longer dictation sessions where holding a key is inconvenient. You toggle it on, speak as long as you need, and toggle it off when finished. Because Clovy operates at the system level, there is no per-app setup. It types into the active text field regardless of which app is open: your email client, a browser text box, a Markdown editor, a commit message in the terminal. The [Clovy private dictation](https://opensoftware.co/clovy/private-dictation-for-mac) feature requires no integrations or extensions to function across apps. --- ## Accuracy and cleanup expectations Clovy is not designed to produce raw transcription. The goal is writing quality output: punctuation is added, filler words are removed, and the result reads like composed text rather than a spoken stream. A few practical habits improve results: - Speak at a measured pace. Very fast speech with no pauses increases transcription errors before cleanup. - Pronounce names, acronyms, and numbers clearly. For unusual proper nouns, say them slowly the first time. - Pause briefly between distinct ideas rather than running sentences together. - Use a decent microphone. The built-in Mac microphone works, but a headset or external mic noticeably reduces errors in noisy environments. Common dictation pain points and how Clovy handles them: - *Filler words (um, uh, like):* Removed by default during the cleanup pass. - *Missing punctuation:* Added contextually based on sentence structure and natural pauses. - *Formatting inconsistencies:* The writing style setting (standard, casual lowercase, formal) gives you a consistent output register across sessions. - *Long dictations:* Clovy supports extended recording sessions for stretches where a single short press-and-release is not enough. --- ## Privacy: what leaves your Mac and what stays local Most dictation tools are vague about this. Clovy is precise. When you dictate, the audio leaves your Mac and is sent for transcription and cleanup. By default, that request routes to zero-retention Venice models. Zero retention means the provider does not store the audio or transcript after the request completes, and it does not use your input for model training. What stays on your Mac: - Your dictation history - Meeting recordings, transcripts, and notes - App state, sessions, and agent memory What OpenSoftware stores on its servers: - Account, login, and billing records - Anything you deliberately submit in an in-app issue report, which sends your description, diagnostics, and attachments to the team Clovy is partially offline. Your dictation history and stored notes remain readable without an internet connection. Dictation itself, note generation, and the local agent all require a live model call, which needs internet. Claiming otherwise would be inaccurate, and [the FAQ](https://opensoftware.co/clovy/faq) documents this directly. Two separate assurances back the privacy claims, and it is worth keeping them distinct. First, the Clovy API runs in a TEE and serves cryptographic attestation at [opensoftware.co/verify](https://opensoftware.co/verify), which lets you verify the code executing inside the enclave. Second, the default Venice routing carries a contractual zero-retention commitment from the provider. The attestation proves what Clovy's own infrastructure runs; the zero-retention commitment covers what the model provider does with what it receives. Together they replace "trust our policy" with something you can actually check. --- ## Clovy compared to the main Mac voice-to-text options in 2026 | App | Category | Privacy default | Offline support | Notes | |---|---|---|---|---| | Clovy | Live dictation + meeting notes | Zero-retention Venice routing by default; TEE-attested Clovy API | Partial (history local; model calls need internet) | Hobby free + $20/month Pro | | Apple Dictation | Live dictation | On-device processing available for general text dictation, depending on settings | Yes (on-device mode, general text dictation) | Free; limited AI cleanup | | Wispr Flow | Live dictation | Cloud only; audio and transcripts stored and usable for training until Privacy Mode is on and Cloud Sync is off | No | Free 2,000 words/week on desktop; Pro $15/month | | Superwhisper | Live dictation + manual meeting mode | Configurable; fully on-device with Pro on Apple Silicon | Yes (Pro, Apple Silicon) | Free tier; Pro $8.49/month; also Windows 10+ with feature gaps | | MacWhisper | File transcription | Locally run models by default; optional cloud providers | Yes (local models) | Not live dictation; version 14 as of July 2026 | | Otter.ai | Meeting notes | Cloud storage; de-identified training on by default | No | OtterPilot bot can join calls; bot-free desktop capture also available | | ChatGPT Desktop | Record mode (meetings) | OpenAI servers; raw audio deleted after transcription, results stored in ChatGPT history | No | macOS 14+, Apple Silicon only; voice mode retired from macOS app January 2026 | A few comparisons worth expanding: **Clovy vs MacWhisper.** MacWhisper transcribes audio and video files you already have, using locally run models like Whisper and Parakeet, with optional cloud transcription providers. It is strong for processing recorded content on your Mac. Clovy types into your cursor while you speak. These are different workflows. MacWhisper 14 (July 2026) added a new transcript editor and speed improvements. If you need to transcribe a recorded interview, MacWhisper is the better fit. If you need to write by voice in your email client right now, Clovy is. You can read a full breakdown in the [Clovy vs Superwhisper comparison](https://opensoftware.co/clovy/vs/superwhisper) for similar tradeoffs around local model modes. **Clovy vs Wispr Flow.** Wispr Flow's transcription always happens in the cloud. By default, audio and transcripts are stored on its servers and may be used to train its models; reaching its zero-retention state requires turning Privacy Mode on and Cloud Sync off, a combination that is only the default on Enterprise and HIPAA plans. Clovy routes to zero-retention models by default on every plan and keeps history on your Mac. The [Clovy vs Wispr Flow page](https://opensoftware.co/clovy/vs/wispr-flow) covers the data control differences in detail. **Clovy vs Superwhisper.** Superwhisper offers a genuine fully on-device configuration: with Superwhisper Pro on Apple Silicon, both the voice model and the language model can run locally, so dictation in that setup never touches a server. Clovy does not offer an on-device model mode. That is an honest tradeoff, and Superwhisper deserves the credit. Superwhisper also has a free tier, a manual meeting recording mode with summaries, and a Windows 10+ version with feature gaps. Outside the on-device configuration, Clovy's privacy defaults are stronger: zero-retention routing by default on every plan, including free, with no manual setup. **Clovy vs Otter.ai.** Otter's OtterPilot bot can join your Zoom, Meet, or Teams calls as a visible participant via calendar integration, and Otter also offers bot-free desktop capture from system audio. Either way, recordings and transcripts are stored in Otter's US AWS cloud, and its trust page says de-identified recordings and transcripts train its models automatically. [Clovy's meeting notes](https://opensoftware.co/clovy/ai-meeting-notes-without-a-bot) record from your Mac's audio directly, so nothing joins the call, and transcripts stay on your device. OpenSoftware never trains on your data. **Clovy vs ChatGPT Desktop.** ChatGPT Desktop requires Apple Silicon and macOS 14+. OpenAI retired voice mode from the macOS app in January 2026. The current Record mode, available on paid plans with a 240-minute cap, captures meetings from system audio and deletes raw audio after transcription, but that processing happens on OpenAI's servers and the results live in your ChatGPT history there. Clovy also requires macOS 14+ and runs on both Apple Silicon and Intel. The [Clovy vs ChatGPT Desktop comparison](https://opensoftware.co/clovy/vs/chatgpt-desktop) documents the practical differences. **Clovy vs Apple Dictation.** Apple's built-in Dictation is free, and Apple documents that general text dictation can be processed on your device rather than sent to Siri servers, depending on your settings and Mac model; you can check this in Keyboard settings. It does not apply AI writing cleanup. Clovy adds punctuation correction, filler-word removal, and selectable writing styles on top of transcription. If you want raw speech-to-text with no cost and no account, Apple Dictation is a reasonable baseline. If you want writing quality output and private AI infrastructure, Clovy is the upgrade. --- ## Dragon for Mac in 2026: discontinued, not replaced Nuance Dragon for Mac is no longer available. Nuance discontinued Dragon Professional Individual for Mac in October 2018, and its own support pages document the end-of-life dates; existing perpetual licenses kept working but the product received no further updates and is not supported on current macOS versions. If you are searching for Dragon Professional for Mac or wondering what replaced it, the short answer is that no single app occupies the same space. For live push-to-talk dictation with AI writing cleanup, Clovy covers that use case. For local Whisper-based transcription of files, MacWhisper covers that ground. For voice navigation and accessibility-focused input, Apple's built-in Voice Control and accessibility tools are worth evaluating. --- ## Pricing and hardware requirements Clovy runs on macOS 14 or later on both Apple Silicon and Intel Macs. - **Hobby (free):** Includes dictation with the same zero-retention privacy standard as Pro. Good for trying Clovy on real work before committing. - **Pro ($20/month):** More usage across dictation, meeting notes, and the local agent. Extended agent sessions, access to all models, and scheduled routines for recurring tasks. Note on offline use: Clovy's history, notes, and files stay on your Mac and remain accessible without internet. Dictation, note generation, and the agent require a live model call and will not function without a connection. The app does not claim otherwise. --- ## Frequently asked questions **Is Clovy's dictation processed on my Mac?** The app and your dictation history live on your Mac. The audio itself is sent out for transcription and cleanup, routed to zero-retention Venice models by default. That means nothing is stored by the provider and nothing is trained on. History stays on your device, not on OpenSoftware servers. **Does it work in every app?** Yes. Clovy operates at the system level and types polished text into whichever app has focus: email clients, browsers, Markdown editors, terminals, chat apps, code editors. There is no per-app configuration. **Does Clovy train on my dictation?** No. OpenSoftware never trains on your data. By default, model calls route to zero-retention Venice models on the Private tier, which store nothing and train on nothing. An Anonymous tier of third-party models is available but opt-in only, and those providers may retain what they receive; Clovy strips identifying metadata before routing. A separate E2EE model tier encrypts your prompt on device so it is decrypted only inside the enclave. **Can I use Clovy without sending data to the internet?** Partially. Your dictation history, notes, transcripts, and files stay on your Mac and are readable offline. Dictation itself requires a model call, which needs internet. Clovy documents this clearly rather than claiming fully offline operation. **Does Clovy transcribe existing audio or video files?** Clovy is primarily a live dictation tool. It also generates meeting notes from live calls recorded through your Mac's audio. For batch transcription of existing audio or video files, MacWhisper is designed for that use case. **What Macs does it run on?** Clovy requires macOS 14 or later and runs on both Apple Silicon and Intel Macs. **Is Dragon for Mac still available?** No. Nuance discontinued Dragon for Mac in 2018 and it is not supported on current macOS versions. Clovy covers live dictation with AI cleanup; MacWhisper covers local file transcription. **How can I verify Clovy's privacy claims?** Visit [opensoftware.co/verify](https://opensoftware.co/verify). The Clovy API runs in a TEE and serves live cryptographic attestation. Clovy is also open source and MIT-licensed, so the full codebase is auditable. --- ## Sources Claims about other products come from their own public pages or named third-party sources, checked on July 16, 2026. - Clovy privacy details: storage, model routing, and retention claims - Clovy source code: MIT-licensed app and Clovy API - [Clovy API attestation](https://opensoftware.co/verify): live TEE attestation you can check yourself - [OpenSoftware privacy policy](https://opensoftware.co/legal/privacy): what OpenSoftware stores - Apple Support, "Use Dictation on Mac": on-device processing for general text dictation, depending on settings - MacWhisper (macwhisper.com): local model transcription with optional cloud providers - 9to5Mac, July 14, 2026: MacWhisper 14 launches with new transcript editor and faster performance - Nuance support (nuance.custhelp.com): Dragon Professional for Mac 6 end-of-life dates; MacRumors, October 24, 2018: Nuance discontinues Dragon for Mac - Wispr Flow data controls page (wisprflow.ai/data-controls) and pricing page: cloud-only transcription, Privacy Mode and Cloud Sync defaults, free tier limits - Superwhisper docs (superwhisper.com): local model requirements, cloud API zero-retention agreements, meeting mode, platforms, pricing - Otter.ai privacy and security pages: storage, training defaults, capture modes - OpenAI Help Center: ChatGPT macOS app requirements, voice mode retirement, Record mode --- ## More guides and comparisons - [AI meeting notes without a bot](https://opensoftware.co/clovy/ai-meeting-notes-without-a-bot): Clovy writes structured meeting notes from your Mac's audio. No bot joins the call, and recordings, transcripts, and notes are stored on your Mac. - [Local AI agent for Mac](https://opensoftware.co/clovy/local-ai-agent): Clovy's agent runs on your Mac, not a company's cloud VM. Open-source Hermes foundation, sandboxed by default, zero-retention model routing. - [Clovy vs Superwhisper](https://opensoftware.co/clovy/vs/superwhisper): Superwhisper has a genuine fully on-device Pro mode; Clovy is private by default with zero setup. An honest comparison of two privacy-minded Mac dictation apps. - [Clovy vs Wispr Flow](https://opensoftware.co/clovy/vs/wispr-flow): Wispr Flow transcribes in the cloud and can train on your data until you change two settings. Clovy routes dictation to zero-retention models by default and keeps history on your Mac. - [Clovy vs Otter.ai](https://opensoftware.co/clovy/vs/otter): Clovy captures from your Mac's audio, keeps transcripts on your device, and never trains on your data. Otter stores recordings in the cloud and trains on de-identified data by default, whether captured by its bot or its desktop app. - [Clovy vs ChatGPT Desktop](https://opensoftware.co/clovy/vs/chatgpt-desktop): A practical comparison of dictation, privacy defaults, and hardware requirements. - [Clovy FAQ](https://opensoftware.co/clovy/faq): Common questions about Clovy, privacy, and pricing. --- ## Try Clovy on real work Clovy is a fit if you want live dictation into any Mac app, AI writing cleanup with selectable styles, and privacy infrastructure you can verify rather than take on trust. Free to start. macOS 14 or later, Apple Silicon and Intel. [Download Clovy](https://opensoftware.co/clovy) --- # Superwhisper vs Wispr Flow: which dictation app is better? > An honest 2026 comparison of Superwhisper and Wispr Flow: processing location, privacy controls, pricing, and platforms, plus a private third option. - Canonical URL: https://www.opensoftware.co/blog/superwhisper-vs-wispr-flow - Published: Jul 16, 2026 - Updated: Jul 16, 2026 - Category: Guides - Author: OpenSoftware - Reading time: 9 min read - Topics: Superwhisper vs Wispr Flow, Mac dictation app, voice to text Mac, dictation privacy, Clovy dictation ## Primary links - [Learn more about Clovy](https://www.opensoftware.co/clovy) - [Download Clovy](https://www.opensoftware.co/download/mac) - [Clovy community](https://t.me/+B4Z8KUqEsRc4ZGVh) - [Open source repository](https://github.com/open-software-network/os-clovy) - [Venice AI](https://venice.ai/) - [Hermes Agent](https://hermes-agent.nousresearch.com/) - [Hermes Agent source](https://github.com/NousResearch/hermes-agent) Both apps solve the same core problem: you speak, text appears. Where they split is *how* that speech gets processed, what happens to the audio afterward, and what you pay when your usage climbs. This guide breaks down those differences across five decision pillars so you can pick the right one for your workflow. ## The short answer Superwhisper is the better pick if you want on-device processing and offline capability. Its fully local mode keeps audio on your Mac entirely, and its cloud models are covered by named zero data retention agreements with the API providers. Wispr Flow is the better pick if you want polished output across multiple devices and don't mind that transcription always runs in the cloud. Neither answer is universal. Here's the map: - **Privacy-first / sensitive dictation:** Superwhisper's fully on-device setup (a Pro feature on Apple Silicon) keeps audio off the network entirely - **Polished cross-device output:** Wispr Flow's cloud pipeline delivers formatted results across Mac, Windows, and iPhone with less setup - **Heavy daily dictation on a budget:** Wispr Flow's free plan caps at 2,000 words per week on desktop (1,000 on iPhone), so moderate-to-heavy users hit the wall quickly - **Offline or low-connectivity environments:** Superwhisper with a local model works without an internet connection; Wispr Flow requires one because transcription always occurs in the cloud - **Meetings:** Superwhisper has a manual meeting mode with summaries and no bot; Wispr Flow does not document a meeting capture feature ## Quick comparison table | | Superwhisper | Wispr Flow | |---|---|---| | Free tier | Yes | 2,000 words/week desktop, 1,000/week iPhone | | Pro pricing | $8.49/mo, $84.99/yr, or $249.99 lifetime | $15/mo, or $12/mo billed annually | | Processing location | On-device (local models, Pro) or cloud APIs with ZDR agreements | Cloud only | | Offline mode | Yes (local model, Pro, Apple Silicon) | No | | Privacy posture | Local mode sends no audio to a server; cloud API usage has ZDR agreements | Zero retention requires Privacy Mode on and Cloud Sync off | | Platforms | macOS, Windows 10+ (feature gaps), iOS 18+ keyboard | macOS, Windows, iPhone | | Meeting capture | Manual meeting mode with summaries, no bot | Not documented | ## Audio processing: local vs cloud and why it matters Superwhisper can run speech recognition on-device using local AI models. With a local model selected, audio never leaves your Mac, which means no network dependency and no third-party server ever receives your audio. Two caveats keep this honest: the fully local setup (local speech-to-text plus a local language model) requires the Pro plan and runs well only on Apple Silicon, and Superwhisper also offers cloud models. For those cloud models, Superwhisper documents zero data retention agreements with its API providers (Anthropic, OpenAI, Deepgram, Groq), scoped to API usage. Its privacy policy is dated June 2024, which predates the cloud model expansion, so read the security docs alongside it. Wispr Flow takes a different architecture. Its own data-controls page states that transcription always occurs in the cloud. There is no local or offline mode. The tradeoff is straightforward: cloud-only processing versus an available on-device option is the core structural difference between these two apps. For latency, neither vendor publishes independent benchmarks. As a practical test, dictate the same 150-word paragraph three times on each app, measure time from end of speech to fully inserted text, and compare punctuation, filler removal, and word accuracy against the source. That gives you a repeatable baseline the vendor pages don't provide. ## Privacy controls: what each app actually promises Superwhisper's position has two parts. In local mode, no audio or transcript reaches a server at all. For cloud model usage, named zero data retention agreements cover the API calls. One thing to know: recordings accumulate on your disk with no auto-delete, so local storage is your responsibility to manage. Wispr Flow's controls are layered. The data-controls page documents two settings: - **Privacy Mode:** when on, prevents data retention and training usage - **Cloud Sync:** when on, syncs transcripts across devices via Wispr's servers Zero data retention on Wispr Flow requires a specific configuration: Privacy Mode on and Cloud Sync off. That combination is the default only on Enterprise and HIPAA plans. On standard accounts you have to set it yourself, and until you do, audio and transcripts are stored on US servers and may be used to train Wispr's models. **What to check before you commit:** 1. Is Privacy Mode on? Is Cloud Sync off? Both are required for Wispr Flow's zero-retention state 2. Where are transcripts stored locally? Can you find and delete them? 3. Does the app use your dictations to improve models if you don't opt out? 4. Is there a DPA or SOC 2 report available if you need it for enterprise compliance? ## Pricing and real-world cost Wispr Flow's free tier applies a weekly word cap: 2,000 words per week on Mac and Windows and 1,000 words per week on iPhone, per Wispr Flow's own pricing page. For someone dictating emails daily, that cap arrives quickly. Wispr Flow Pro runs $15 per month, or $12 per month billed annually. Superwhisper offers a free plan, Pro at $8.49/month or $84.99/year, and a one-time lifetime purchase at $249.99, per superwhisper.com. Note that the fully on-device configuration (local speech-to-text plus local language models) is a Pro feature, so budget for Pro if local processing is why you're choosing it. **Lifetime break-even math (estimates):** At Wispr Flow Pro's $15/month, you'd spend $249.99 in roughly 17 months. Against Superwhisper's own Pro at $8.49/month, the lifetime plan breaks even in about 29 months. If you plan to use a dictation app for more than two to three years, Superwhisper's lifetime option is the cheaper long-run choice. If you prefer cross-device flexibility or you're unsure about long-term commitment, monthly subscriptions keep options open. ## Platform support and ecosystem fit Superwhisper is macOS-first and works system-wide, meaning you can dictate into any app, including Mail, Slack, Notion, code editors, and browser text fields. It also ships a Windows 10+ version with feature gaps relative to the Mac app, and an iOS 18+ keyboard. Offline models run well only on Apple Silicon, per Superwhisper's own FAQ. Its meeting mode is manual recording with summaries and does not use a bot; speaker separation is a Pro feature. Wispr Flow supports macOS, Windows, and iPhone, which matters if you work across devices. Its cloud architecture is what makes multi-device sync practical, with the privacy tradeoff described above. On macOS it also works system-wide. ## Decision guide by scenario **You need offline dictation or you handle sensitive material (legal, medical, financial):** Superwhisper with a local model is the clear choice. Audio stays on your Mac and processing is on-device. Plan on Pro and Apple Silicon for the fully local setup. **You want polished output with minimal setup and you work across Mac, Windows, and iPhone:** Wispr Flow's cloud pipeline and cross-platform support make it the more practical option, provided you turn Privacy Mode on and Cloud Sync off if data retention matters. **You're a light user on a budget:** Wispr Flow's free tier covers casual use at 2,000 words per week on desktop. Superwhisper also has a free plan; check its current limits on superwhisper.com before committing, and remember the fully local configuration sits behind Pro. **You're a heavy daily dictator planning a multi-year commitment:** Superwhisper's lifetime license at $249.99 costs less than Wispr Flow Pro's subscription after about 17 months. **You want dictation plus meeting notes plus an AI agent in a single private app:** Neither tool covers all three natively. [Clovy by OpenSoftware](https://opensoftware.co/clovy) bundles push-to-talk [private dictation for Mac](https://opensoftware.co/clovy/private-dictation-for-mac) into any app alongside [AI meeting notes without a bot](https://opensoftware.co/clovy/ai-meeting-notes-without-a-bot) and a local agent. Prompts route to zero-retention Venice models by default on the Private tier, the Clovy API runs in a Trusted Execution Environment with cryptographic attestation, and the codebase is MIT-licensed so the privacy claims are verifiable rather than just stated. It requires macOS 14+. The Hobby tier is free to start; Pro is $20/month. ## Setup and verification checklist Before you rely on either app for sensitive dictation, take five minutes to verify the settings actually match the privacy guarantees. **For Superwhisper:** - Open Settings and confirm a local model is selected, not a cloud API (fully local setup requires Pro on Apple Silicon) - Dictate a short clip, then watch for network activity (use Little Snitch or macOS Activity Monitor's network tab) - Review the privacy policy and security docs at superwhisper.com; the privacy policy is dated June 2024 and predates the cloud model expansion, so check the security documentation for current cloud handling - Remember recordings accumulate on disk with no auto-delete; clear them on your own schedule **For Wispr Flow:** - Open Settings and confirm Privacy Mode is on and Cloud Sync is off; both settings together are what Wispr documents as zero data retention - Remember transcription still runs in the cloud even in this configuration; zero retention means Wispr doesn't keep it, not that it never leaves your device - If you need a DPA or SOC 2 report, contact their team directly ## FAQ **Which is better for Mac-only users?** Both work well on macOS. Superwhisper's offline mode and local processing give it an edge for privacy-sensitive Mac workflows. Wispr Flow's polish and cross-device sync matter more if you also use Windows or an iPhone. **Does either app require an internet connection?** Superwhisper does not when you use a local model. Wispr Flow requires an internet connection because transcription always occurs in the cloud. **Can I use it for meetings without a bot joining the call?** Superwhisper has a manual meeting mode with summaries and no bot joins the call; speaker separation is a Pro feature. Wispr Flow does not document a meeting capture feature. See [Clovy vs Wispr Flow](https://opensoftware.co/clovy/vs/wispr-flow) and [Clovy vs Superwhisper](https://opensoftware.co/clovy/vs/superwhisper) for how each compares to a privacy-first alternative with built-in bot-free meeting notes. **Is my dictation used to train models?** Superwhisper's local mode keeps audio and transcripts off servers entirely, and its cloud API usage is covered by zero data retention agreements. Wispr Flow may use stored audio and transcripts to train its models until you turn Privacy Mode on and Cloud Sync off; that configuration is the default only on Enterprise and HIPAA plans. **How do Wispr Flow's free-tier word caps work?** The cap resets weekly. You get 2,000 words per week on Mac and Windows and 1,000 words per week on iPhone, per wisprflow.ai/pricing. Exceed the cap and you'll need to upgrade to Pro ($15/month, or $12/month billed annually) or wait for the next reset. **What happens to your data if you turn off Privacy Mode in Wispr Flow?** Based on Wispr Flow's data-controls documentation, with Privacy Mode off your audio and transcripts may be retained on Wispr Flow's servers and used for model improvement. If you want stored data removed, contact Wispr Flow's support for their deletion procedure.