What we collect, why, how to delete it.

The service is a chat interface over GitHub repositories. It packs a repo's source code and lets you ask questions about it. This page describes everything we collect, including the conversations themselves and how we handle GitHub access.

What we store about your account

When you sign in with GitHub, we store:

Your GitHub login, numeric ID, and avatar URL, used to identify your account.
Your IP address at sign-in, used to enforce per-IP abuse limits (multi-account creation from one network triggers a 24-hour cooldown).
Your daily and weekly chat-message counts, used to enforce your plan's message limits.
A session cookie (random token, 30-day expiry), used to keep you signed in.
Your consent timestamp, recording when you agreed to the Terms and chat-logging at sign-in (the audit trail required by GDPR Article 7).
Your GitHub access + refresh tokens, encrypted at rest — see the next section.
Your subscription status, if you activate a paid plan — the plan tier, its status, and a one-way hash of your license key, tied to your GitHub ID. We never store your card details (Gumroad handles payment).

We do not store passwords (sign-in is GitHub OAuth only). We do not run third-party trackers, analytics, or advertising scripts.

GitHub App access and tokens

Sign-in is handled by our GitHub App. The permissions it requests are read-only: repository Contents and Metadata — enough to read the code of repositories you grant access to, and nothing else (no write access, no issues, no actions, no organization administration).

The access token (short-lived) and refresh token GitHub issues are encrypted at rest in our database and used only to read repository content on your behalf when you chat.
You can revoke the service's access at any time at github.com/settings/apps/authorizations. Revoking there immediately cuts off our ability to read your repos.

Private-repo code

Plain-English summary: When you chat about a private repo, we read its code, hold it in memory just long enough to answer, and never write that code to our database or cache.

For private repositories, the packed source code is processed in memory only, for the duration of the request, and never stored. The system prompt (which contains the packed code) is rebuilt on every request for private repos rather than persisted, and the packed context is never written to our cache. Only the chat messages and replies are logged, per your consent, and only if you leave logging on.

Conversation logging

For public-repo chats, we save the conversation so you can resume it later and so we can debug and improve answers. For each chat we store:

The repo you're chatting about (e.g. facebook/react) and whether it is private.
The full text of every message you send, with a timestamp.
The full text of every AI reply, with the model version that produced it.
For public repos only, the packed source context used as the AI's initial context (cached so follow-up turns are consistent). This is never stored for private repos.

Anonymous chats (the free messages before signing in): we also log these, attached to a short-lived anonymous cookie. They aren't linked to any account.

Why we log: debugging bad answers, improving the system prompt, and building a per-account history so you can resume past chats. We do not sell or share these logs.

Revoking consent and deleting chats

Revoke logging: from /me/history, click "Turn off logging." Future chats won't be saved server-side. Past chats remain readable until you explicitly delete them.
Delete a single chat: open it from history and click "Delete chat." Both the messages and the chat metadata are permanently removed.
Delete account + all chats: email [email protected] from the address tied to your GitHub account. We'll confirm full deletion within 7 days. Revoke the GitHub App at github.com/settings/apps/authorizations too.

Processors we rely on

Cloudflare (US) — hosting, edge delivery, and our databases (D1) and caches (KV). Cloudflare's privacy policy applies.
OpenRouter (US) — routes your chat messages, the repo context, and your prior turns to the language model that generates replies. OpenRouter's privacy policy applies.
GitHub — the source of the repository data and the OAuth identity. GitHub's privacy statement applies.
Gumroad (US) — Gumroad, Inc. processes payments and subscriptions for paid plans as merchant of record. When you buy Pro or Max, Gumroad receives your email and payment details (we never see your card) and issues the license key you activate. Gumroad's privacy policy applies.

Bot detection

The anonymous chat path includes a Cloudflare Turnstile widget — an invisible bot-detection layer that runs in your browser without storing personal data. Cloudflare's privacy policy applies to that interaction.

Legal basis

For users in the EU/UK, the lawful basis for processing your account and chat data is consent (UK GDPR Art. 6(1)(a) / EU GDPR Art. 6(1)(a)). Consent is given when you sign in and agree to the Terms and chat-logging. You can withdraw it at any time via "Turn off logging" on the history page — see Revoking consent. For a paid subscription, the lawful basis for processing your billing and entitlement data is performance of a contract (Art. 6(1)(b)).

Children

The service is not directed at children under 16. If you believe a child has signed up, email [email protected] and we'll delete the account.

Changes to this policy

Material changes will be reflected here with the "Updated" date above. Non-material changes (typos, clarifications) are made in place.

Contact

BitVibe Labs · [email protected]