Last updated 17 September 2026

Privacy

Rivet reads your source code, runs it somewhere else, and sends parts of it to a model provider. That is not a footnote, so it is the first thing this page says. Below is exactly what is read, where it goes, what is kept, and how to get rid of it.

Who is responsible

Rivet is operated by [LEGAL ENTITY NAME], at [REGISTERED ADDRESS]. Questions about this notice, or about your data, go to [email protected].

The bracketed items above are gaps the operator has to fill in. They are marked rather than invented.

What we collect

Three kinds of thing, and nothing else.

  • Your account. Your GitHub login, your name and email address as GitHub reports them, and the access tokens needed to act on repositories you have installed the app on.
  • Repository content. Manifests, lockfiles and source files from the repositories you connect, plus issues, comments and Dependabot alerts on them. A scan reads the file tree; a run downloads an archive of the default branch.
  • Run records. What each run did: the packages and advisories involved, the commands executed, truncated test and command output, the diffs produced, excerpts of the pages consulted, the model calls made, the tokens used and the cost.

We do not use analytics or advertising trackers, do not sell anything to anyone, and do not use your code or your run records to train models.

Where your code goes

A run cannot happen inside this website. Your repository is checked out into a third-party sandbox, and excerpts of it are sent to a third-party model provider. These are the processors involved:

Nebius Token Factory (Sandboxes)

Runs a copy of your repository: installs dependencies, applies the upgrade, runs your tests.

Repository source, lockfiles and test output for the duration of a run.

Nebius Token Factory (inference)

Serves the NVIDIA Nemotron models that plan the upgrade, edit the code and review the diff.

Excerpts of your code, test failures, advisory text and changelog text.

GitHub

Where your repositories live. Rivet reads them and writes a branch and a draft pull request.

Repository content, issues, comments, Dependabot alerts.

Tavily

Searches for and extracts the changelog and migration guide for an upgrade.

Package names and version numbers. Never your code.

OSV.dev and the GitHub Advisory Database

The advisory data itself. Queried by package name and version.

Package names and version numbers. Never your code.

Sandbox and inference run on Nebius infrastructure. Where a run physically executes depends on the region the platform assigns; the operator does not choose it per run. [CONFIRM PROCESSING REGIONS BEFORE RELYING ON THIS FOR GDPR TRANSFER TERMS]

Your GitHub credentials and your repository secrets are never put inside the sandbox. The sandbox gets a copy of the code; the tokens stay outside it.

What is kept, and for how long

  • Sandboxes are torn down when a run ends. The checkout does not persist between runs.
  • Run records persist in our database so you can read what happened afterwards, including the diff and excerpts of source files, test output and pages consulted. They are kept until you delete the run or your account.
  • Account and installation data is kept while your account exists.
  • Public scans from the scanner on this site are cached briefly so a popular repository is not rescanned on every visit. They are not attached to an account, and a scan never writes to a repository.

[SET A RETENTION PERIOD FOR RUN RECORDS ON DELETED ACCOUNTS] Until one is set, deletion on request is the mechanism below.

Deleting your data

  • Uninstall the GitHub App to stop all reading and all writing immediately. Nothing runs after that.
  • Pause a repository from the dashboard to stop new runs on it without uninstalling anything.
  • Email [email protected] to have your account and every run belonging to it deleted. We will confirm when it is done.

Pull requests Rivet has already opened belong to your repository, and deleting your Rivet account does not touch them. Close them on GitHub if you do not want them.

Your rights

Depending on where you live you may have rights to access, correct, export or erase your personal data, and to object to processing. Exercise any of them by emailing [email protected]. The supervisory authority you can complain to depends on your country; ours is [SUPERVISORY AUTHORITY / JURISDICTION].

Changes

If this notice changes in a way that affects what is processed or where it goes, the date at the top changes. Rivet is early and the pipeline is still moving, so check this page rather than assuming it is the version you read last. Related: Terms.