Legal

Privacy

Read Terms instead

Draft — not in force. This text has not been signed by the provider, and 1 part is still to be filled in (marked below). Nothing here binds anyone until the signed version replaces it.

Privacy at Atrytone

Atrytone looks after apps built with AI: it reads your app's code, says in plain words what is wrong, and changes nothing without your yes. This page says what that means for your data, in plain words. The technical note (https://atrytone.ai/next/privacy/technical) backs every line here with the mechanism that makes it true, for anyone who wants to check rather than trust.

Who is responsible for your data: [OPERATOR: the provider's name and postal address]. Atrytone is run by one person; there is no data protection officer. Questions and requests: hello@atrytone.ai, or "Write to us" inside Atrytone — a person answers within one working day, and acts on a request about your data within one month.

The short version

  • We read your app's code only through the GitHub App you install, or the code you upload yourself.
  • Your app's AI work runs on the AI provider you chose, with your own key. We never run it on a model we pay for.
  • We never use your code or your data to train an AI model, and we never sell it or use it for advertising.
  • We publish nothing of yours.
  • Everything Atrytone does for you is written into a permanent record that you can read and download.
  • You can stop us at any time, take your data with you, and ask us to delete your workspace.
  • Atrytone is for people aged 18 and over.

What we collect, and why

  • Your account: your e-mail address and password, to sign you in. Our sign-in service also records when you sign in, with the network address and browser identification of each sign-in, to keep accounts safe. We need your e-mail address to give you an account; without it you cannot use Atrytone.
  • An invitation request: the e-mail address you give, where on the site you asked, a scrambled form (a hash) of your network address, and your browser's identification — to answer your request.
  • Your GitHub connection: your GitHub account name and the names of the repositories you connect, so we know which code is yours.
  • Your code: read from the repositories you connect (or the code you upload), to check it every night, to plan a fix or a feature, and to make and test the change. The nightly check keeps only what it found, never the code, and never the value of a secret it found. A job works on a copy of your code on our server, which is deleted when the job ends. A job keeps a snapshot of the files it read and wrote, with the job, so that a stopped job can carry on. The change itself goes to your own repository, as a pull request.
  • Secrets in your code: files that look like secrets (such as .env or key files) are never shown to the AI. They are on our server with the job's copy of your code, because your installer and your tests may need them, and go when that copy goes. If your tests print a secret, the job's log carries it, and the AI that checks the work reads that log. A secret written inside ordinary code cannot be told apart from code — so please do not commit one.
  • What you ask for: the words you type when you ask for a fix, a feature or an app, your answers to a job's questions, and the conditions you sign — to do the work. Our server's own log also records them while a job runs.
  • The record of what happened: every act — what you signed, what was allowed or refused and why, what each AI call cost, what was checked, what was merged.
  • Files a job makes or you give it: a job's transcript, screenshots, pictures you upload, kept in private storage that only your workspace's members can open, one file at a time.
  • Messages you write to us: with your e-mail address and the page you wrote from, so we can answer. A copy is e-mailed to our own mailbox.
  • E-mails we send you: when something about your apps needs you. Every such e-mail has a link to turn them off.
  • Server logs: a network address, time, the page asked for and your browser's identification, to keep the service running and safe.

Where it is kept

  • Our server is in Vienna, Austria (netcup). A job's copy of your code, and code you upload instead of connecting GitHub, are kept there.
  • Our database, sign-in and file storage are in Frankfurt, Germany (Supabase).
  • The same server and database also serve our other product, Siire, whose data is kept apart from yours.

Who else receives it

  • The AI provider you chose, with your own key — your code and your requests go to it under its terms, not ours. Anthropic, OpenAI and Google are in the United States; that transfer is made under your own agreement with your provider.
  • GitHub — your repository, through the App you install. The App can read your code and write branches, pull requests and check results. It cannot touch your repository's administration, secrets, actions, workflows or members.
  • Supabase — our database, sign-in and file storage.
  • Resend — sends our e-mails: the sign-up confirmation and password-reset e-mails, e-mails about your apps, and the copy of a message you write to us. It receives your address and what the e-mail says.
  • osv.dev, the public list of known flaws in open-source parts (run by Google) — the names and versions of the parts your app uses, never your code.
  • ImprovMX and Google (Gmail) — mail you send us, and the copies of messages you write to us, are forwarded to our inbox.
  • Telegram (and ntfy.sh if Telegram fails) — short notices to the person who runs Atrytone, such as "a customer wrote" or a job's number. They never carry your own words or your address.
  • Tailscale — carries GitHub's pull-request notices to our server.
  • Vercel — answers the address lookups for atrytone.ai. None of your content passes through it.
  • Tools you connect yourself (on the Tools page, such as Supabase or Stripe): Atrytone calls them for you, with the access you granted there, and what a call sends goes to that service. An AI client you connect with a personal token reads your workspace only, and sends what it reads wherever that client sends it.
  • Anthropic, for our own administration. This is separate from your app's AI work. Atrytone is run by one person, who can read every record in the database, as can the tools they run. That person works with an AI assistant, Claude, made by Anthropic. By default, what they give it to read about a workspace is counts, ids, statuses and health — never that workspace's own content (its code, the detail of its findings, your messages or your files). It reads a workspace's content only for a support case that workspace opened, or during an incident, and never without a record: before such a read, we write a permanent entry naming the workspace, which of those two reasons it is, and what will be read. (Open item — see "What is still open before this is signed" below.)

Outside the EU

Our server (Austria), our database (Germany) and our backups are in the EU. Some of the companies above are based outside it: GitHub, Cloudflare, Resend, Google and Anthropic in the United States; Supabase is a United States company that keeps our data in Frankfurt; Telegram is outside the EU and receives no personal data about you. Where a company outside the EU receives personal data from us, the transfer rests on the European Commission's adequacy decision for the EU-US Data Privacy Framework where that company is certified under it, and otherwise on the Commission's standard contractual clauses in that company's data processing terms. Ask us and we tell you which applies to which company, and send you a copy of the clauses.

How long we keep it

  • Your account: until you ask us to delete it.
  • Your workspace — apps, jobs, their snapshots, stored keys, files and the messages you wrote to us: while your workspace exists.
  • A job's copy of your code on our server: deleted when the job ends. A daily check removes any copy left behind once it is 14 days old.
  • A request to be let in: 12 months, then deleted.
  • An account whose e-mail is never confirmed: 30 days, then deleted.
  • An invitation nobody uses: 12 months, then deleted.
  • Server logs: at most 30 days. Since 7 October 2026 our web server's log shortens every network address (it drops the last part) and leaves out sign-in codes.
  • Mail you send us, and our answers: up to two years.
  • Copies: the database is copied once a day; we keep the last fourteen copies, and Supabase keeps its own for about a week. A deletion reaches a copy when that copy ages out. Every deletion you ask for is written down, by number, never by address. If we ever restore a copy, we redo every such deletion made after that copy was taken, and run the time-based deletions above again.
  • The permanent record is never edited. A correction is a new entry. When a workspace is deleted, the record's entries about the work stay: job titles, the conditions you signed, file paths, amounts, verdicts, your GitHub account name and repository names, and the account number of each member — a random number, not your address. They never hold a secret, a password or your e-mail address. We keep them for as long as Atrytone keeps its record, because an intact record is what lets you, us, and anyone you show it to check what was done, and what we would need if the work were ever disputed (GDPR Art. 6(1)(f) and Art. 17(3)(e)).
  • We publish a fingerprint (a hash) of the whole record. It reveals nothing about any entry, and it shows if an entry is ever removed.
  • Everything else of yours — projects, jobs, their snapshots, stored keys, files and your sign-in — is deleted when you ask us to delete your workspace, within one month of your request.

Your choices and your rights

  • Stop: remove the GitHub App from your repository, and nothing can read it again.
  • See it: the Record page shows everything done for you, newest first, and downloads it as CSV or JSON.
  • Take it with you: ask, and we export your workspace — its records and its files.
  • Delete it: press "Ask us to delete my workspace" in Settings, or write to hello@atrytone.ai. A person deletes your workspace and your sign-in, after an export. The permanent record's entries stay, as said above.
  • Ask why: every act has a record, and each finished job has a receipt.
  • Ask, correct, limit: you can ask what we hold about you and get a copy, ask us to correct it, or ask us to limit how we use it while something is checked. Write to hello@atrytone.ai or use "Write to us". We act within one month; if we need longer, we tell you why within that month. If you write from outside Atrytone, we may ask you to confirm from the address you sign in with.
  • Your right to object. Where we rely on our legitimate interests — keeping the record intact, keeping the service safe, and our own administration — you can object at any time, on grounds relating to your situation. We then stop, unless our reasons override yours or we need the data to establish or defend a legal claim, and we tell you which.
  • No automated decisions about you. The AI in Atrytone judges code, not people; nothing decides about you by automated means alone.
  • You can complain to the data protection authority where you live or work. Ours is the Estonian Data Protection Inspectorate (Andmekaitse Inspektsioon, https://www.aki.ee).

Your browser

We set no cookies of our own and run no analytics or tracking. Your browser keeps, on your own device: your sign-in, the app you last chose, a request or a new app you started and did not send, and where to return after you sign in. They are needed for the pages you use, so we do not ask for consent. Our pages load their typefaces from our own server.

Running the service you asked for — your account, your workspace, the work, the record you read, and the e-mails about your apps (a contract, GDPR Art. 6(1)(b)); answering an invitation request you made (steps you asked for before a contract, Art. 6(1)(b)); keeping the record intact after a deletion, keeping the service safe, and our administration (legitimate interests, Art. 6(1)(f)); keeping what the law requires us to keep, such as a record of a security incident (a legal obligation, Art. 6(1)(c)). (Not yet reviewed by a lawyer.)

Children

Atrytone is for people aged 18 and over. If we learn that an account belongs to someone younger, we delete it.

What is still open before this is signed

A few lines above depend on work still in progress, or on a decision only the provider can make. We name them here, plainly, rather than quietly asserting them:

  • Our own AI assistant and training. We use an AI assistant (Claude, by Anthropic) for our own administration, as said above. Whether Anthropic trains its models on what it reads this way depends on which of Anthropic's plans we use for that; we state the answer here, one way or the other, once the provider has confirmed it.
  • How long we keep things. The periods under "How long we keep it" are what our system does today. The provider confirms them when this note is signed.

When this changes

A changed text is a new version. Once this note is signed, a member who accepted an earlier version is asked again.

DRAFT, unsigned. The provider signs it before it is in force.