Privacy Policy

Last updated: September 2026

ReviewDocs is a document-native review experience for Markdown in GitHub repositories. This policy explains what we store and what we don't. It is written to reflect how the product actually works.

What we do NOT store

We do not store your repository contents. Markdown documents, diffs, file trees, commits, and pull request data are fetched from GitHub on demand each time you open a page, rendered in memory, and streamed to your browser. They are not written to our database. The one exception is the optional AI features described below.

What we store

We persist only review metadata needed to power discussions:

• Your GitHub identity (id, login, name, avatar URL, email). If the deployment has email notifications on, we use the email to tell you about replies in threads you're part of; every email has a one-click unsubscribe link.
• Review comments made on GitHub pull requests that change Markdown, and their resolution state, so conversations stay in sync.
• Installations and repository identifiers for repos where ReviewDocs is installed.
• Review threads and comments you write, including short text snippets of the anchored block (so a comment can survive document edits).
• Approval state on pull requests.
• Per-repository settings chosen by repository admins.
• If you create MCP access tokens: a one-way hash of each token (never the token itself), its name, expiry, last-used time, and the installation ids it may read.
• If a repository admin enables guest review links: a one-way hash of each link's token, the pull request it covers, its label, expiry, and who created it; and, for each guest who joins, the name and optional email they enter. A guest's name is shown on their comments (including any mirrored to GitHub); their email is only visible to ReviewDocs.

Optional AI features

AI features (change summaries, plain-language rewrites, and AI review) are off by default and can only be turned on per repository by a repository admin. When they are on and someone uses them:

• The relevant Markdown from that repository is sent to an AI model through the Vercel AI Gateway. Every request requires zero data retention: it is only routed to model providers that have agreed not to retain it or train on it, and is rejected rather than sent anywhere else.
• The AI's output (summaries, rewrites, and review findings, which may quote short passages of your document) is cached in our database so repeat requests are fast and don't re-send your content. AI review findings are saved as review threads.
• This cached output belongs to the repository and is deleted with it when the app is uninstalled.
• A repository admin may also opt in to posting AI findings to the pull request as comments from the ReviewDocs app.

When AI features are off for a repository, none of its content is sent to an AI provider.

Access tokens

Your GitHub OAuth token is stored only in an encrypted, HTTP-only session cookie in your browser — not in our database. The GitHub App's private key and webhook secret live in server environment variables.

Who we share data with

We call the GitHub API to read repository content and to post review comments back to your pull requests. We use a hosting provider and a managed Postgres database to run the service, and — only for repositories with AI features on — an AI provider as described above. We do not sell your data or use third-party advertising or analytics trackers.

Data deletion

When you uninstall the ReviewDocs GitHub App from an account, we delete that installation's repositories and all associated review threads, comments, approvals, settings, guest links, cached AI output, and the records of guests whose comments were removed. You can revoke MCP tokens at any time from your account settings, and guest links from the pull request page. To request deletion of your user record, contact us.

Contact

Questions about privacy? Email support@reviewdocs.ai.