Privacy policy

The desktop app keeps your work on your Mac

Projects are ordinary folders you choose. Accounts, settings, conversations, drafts, Memory and learning notes live in the app's own data folder on your computer. The app sends no usage analytics or crash reports to us. It checks github.com for updates, and its Community features talk to researchtogether.co.

Your OpenAI and Anthropic API keys are encrypted with the macOS keychain. A Codex sign-in is kept by Codex as a file in the app's private data folder, which only your macOS user can open. None of these is sent to us.

What the app sends to AI providers

When you use AI, what the task needs goes to the provider you chose (OpenAI, Anthropic, or OpenAI's Codex) under that provider's own terms:

• your messages, the files and selections you include, and what the assistant read earlier in that conversation;
• your About me and Current focus notes from Memory, with every task;
• the page you are reading as an image, with your marks and notes on it, when a chat works on that page, and to OpenAI when Tutor coaching is on;
• screenshots of the browser or app window, when you let a task use the computer.

When you allow internet access, the assistant can search the web and open pages, so those searches and sites see its requests. When a conversation uses OpenAI, OpenAI keeps each response for 30 days, so later requests can use what its web searches found.

What Community stores when you sign in

Community sign-in uses Google. We keep the account identifier Google returns, the name on the account, which becomes your public name, and your email address. We do not store your Google password. Your email address is never shown on your profile or to other people. You can change your public name, introduction and Now line in your profile.

Email we send

After you first sign in, we send a welcome email and four short tips over the next two weeks. When there is new activity on a data request you posted or responded to, such as a response, a reply or a decision, we email you a summary at most every ten minutes. Every email has a link to stop that kind of email, and you can turn each kind off in Settings. We use your address only for these emails and do not share it with advertisers.

Responding to a request without an account

A request's link lets anyone holding it read that one request and respond. If its poster allows it, you can respond without an account: we email you a six-digit code to confirm your address, and keep your name, your email address, what you send and a cookie that keeps you signed in on that browser for 30 days. We email you about replies and decisions on your response, with a link to stop.

Your email address is not shown to anyone unless you tick the box to share it with your response; then the people in the project can see it. Your name and what you send are visible as the request says, like any response. Files you send are stored on the poster's account.

On the request's page you can delete everything you sent without an account, which also forgets your address. If you confirm your email but send nothing, we forget you 30 days after your last visit.

What you publish and do in Community

We store on our servers the projects you share and the files in the versions you publish, your posts, replies and discussion messages, data requests and responses, the history of changes to shared work with who made each one, your profile name, introduction and Now line, the people and projects you follow, and your likes, saves, blocks, reports and notifications.

New projects are private to the people you invite until you make them public. Public projects, posts and profiles can be read by anyone, including people without an account, and can be downloaded into other people's copies of the app. Who is in a public project is public too: its members are listed on it and on each other's profiles.

When you connect a desktop, your Memory is kept online with your account so your Macs share it; nobody else can open it. The one exception is what you write under a Public profile heading in your About me and Current focus notes: that text is your profile's introduction and Now line, and anyone can read it.

A shared project's work in progress, including text as it is typed, the files in a folder on your Mac you connect to it and who is working in it, is stored on our servers and shown to the project's members. In a public project, anyone can read it and anyone signed in can change it.

Private messages and reports

Direct and group messages are stored on our servers so the people in the conversation can read them. They are not end-to-end encrypted. When someone reports a message, post, project or profile, our operators see what was reported and may hide it or suspend the account. People who have worked on a public project can vote to pause someone's edits there or to freeze their account; we store each vote with who cast it, and our operators review every account frozen this way.

Research in the browser

If you add an OpenAI API key for browser research, it is stored encrypted on our servers and only the last four characters are ever shown back to you. Browser conversations are stored with your account, including what the assistant read with its tools, so a later request can continue from it. OpenAI keeps each response for 30 days, so later requests can use what its web searches found.

ResearchTogether credits

If you pay for AI with ResearchTogether credits instead of your own key:

• Your requests from the Mac app and the browser pass through our servers to OpenAI or Anthropic, which receive them under our account, so their terms and data-retention rules for our account apply. OpenAI keeps each response for 30 days, as above.
• We don't keep what a request contains. We record its provider, model, time and cost to charge your balance and show your history. Browser conversations are stored with your account as described above.
• Stripe handles payment and receives your payment details; we keep each purchase's amount, time and Stripe reference.
• Deleting your account refunds your unused balance to your original payment method, then deletes these records and our Stripe customer record for you. Stripe keeps its own payment records.

The Background agent

If you turn on the Background agent, it runs on our servers on the schedule you choose, using the OpenAI API key you saved for browser research, while your computer can be off. It reads Community and the projects you select, and sends what it reads to OpenAI. Its findings, run history and settings are stored with your account. It posts or replies only after you approve that exact action under Needs you; a request you leave unanswered expires after seven days. If you let it edit projects you selected, its changes apply at once, public projects included, and are recorded in the project's history as your agent's. It cannot reach your computer.

Cookies

The website sets one session cookie after you sign in, valid for 30 days, and a short-lived cookie that lasts ten minutes during the Google sign-in step. If you respond to a request without an account, a separate cookie keeps you signed in to request pages for 30 days. There are no advertising or analytics cookies.

Where data is stored and who processes it

Community data is stored in the United States. The website runs on Vercel, the database is PostgreSQL hosted by Supabase, files are stored in Amazon Web Services S3, sign-in is handled by Google and email is sent through Resend. Each acts as a processor for the data described above. AI providers you connect handle what you send them under their own terms. We do not sell personal data or share it with advertisers.

To limit abuse, the service counts requests by a one-way hash of your network address; those counters expire after a minute, or within a day for email codes and responses without an account, which are also counted by a one-way hash of the email address. Records that stop a retried request from being applied twice keep only a fingerprint of the request, and those for posts and messages are removed after seven days.

Keeping and deleting data

Community data stays until you delete it or ask us to. A public project stays public, because other people may be working on it; you can make a private project public at any time.

The owner of a private project can delete it from People & access, on the web or in the Mac app, by typing its title. That removes it for everyone: its versions, history, files, work in progress, discussions, requests and responses, follows and the research conversations opened in it. A public project can only be removed by us, for example when it breaks the Terms. We keep only its identifier and the time, so its address can say the owner deleted it. Deleting a project never deletes files on anyone's computer, and copies people already downloaded stay with them.

You can delete your own response to a data request at any time, and the owner of a request can delete it with every response, which also turns off its link. Copies others downloaded and results they published cannot be recalled.

To delete your account, email us from the address you sign in with. We then delete your profile, sign-ins and connected desktops, email records, follows, blocks, likes and saves, reports you filed, notifications, browser research conversations and saved key, your Background agent and its history, your responses to requests, and every private project you own. Public projects you own stay public, owned by “Deleted account”. So that other people's replies still make sense, each message you wrote in posts, discussions and conversations with them stays in place as “[Removed]” from “Deleted account”, and posts you started leave the feed. Changes you made to a project that stays are kept in its history, credited to “Deleted account”.

Desktop data is yours to delete by removing the app's data folder and your project folders.

Children and changes

ResearchTogether is not directed at children under 13, and you must be 13 or older to use Community (see the terms). When this policy changes, the new version appears here with a new date.