This page explains what data Procli Code handles — account, repository, scan, session, and API-key data — and how it is used to provide quality and security analysis for software built with AI.
Last updated: July 2026
This policy covers Procli Code: the web dashboard, the Cheat and Guard agent integrations, the Quality and Audit scanners, and the command-line interface (CLI). It describes the specific data each part handles, why, and the controls you have over it.
During the current private beta, access is limited to allowlisted accounts (by email address or account ID). Signing up does not guarantee access while that gate is on.
Sign-in is handled by Clerk, our authentication provider. For your Procli account we store an account ID, your email address, an optional display name, and — if you sign in with Google — the associated Google account identifier, along with your account status and created/updated timestamps.
We do not store your password; Clerk manages credentials and the sign-in session. We use this data to sign you in, gate the dashboard, enforce beta access, and attach the rest of your data to your account.
You connect repositories by installing the Procli GitHub App. We store the installation ID and, for each repository you connect, its owner, name, full name, GitHub repository ID, and default branch. If you disconnect a repository, we record that disconnect so it is no longer scanned.
GitHub installation access tokens are used server-side to read repository contents only when a scan runs, and are not stored in the browser. Access is scoped to the repositories you choose in GitHub, and you can change or remove it at any time from your GitHub App settings.
When you run a Quality or Audit scan, we store a scan run: its type (quality or audit), status and stage, the resulting findings as structured JSON, and any error message. Findings can include file paths, code references, severity, and fix guidance so you can act on them, and are also offered as JSON you can hand to your own agent.
Scan runs carry an expiry, so stored results are cleared after they age out rather than kept indefinitely. Scanning is metered per account (by default up to 10 scans per day, with other actions limited to protect the service).
Cheat and Guard operate on your agent's session. Cheat reads tool-use and session-end events (and commit messages) to check a claim against what actually ran; Guard inspects an action — for example via Claude Code PreToolUse hooks — before it executes. The signals involved are the commands run, files changed, and results returned.
Guard and Cheat findings are owned and stored by the separate Procli engine, not by this web app. The dashboard reaches that engine only through a server-side proxy: each request verifies your identity, resolves your engine tenant server-side, and is signed with a short-lived (about five-minute) key over an encrypted connection — no static shared secret, and no token ever reaches the browser. This web app stores only your Cheat and Guard preferences (such as the rollout stage and visibility toggle) and which capabilities are enabled for your account.
CLI keys carry the prefix procli_. We store only a one-way hash of each key plus enough to display it safely (its prefix and last characters); the full key is shown once at creation and never stored or shown again. We also record when a key was created, last used, and revoked.
Linking the CLI uses a device flow: procli login requests a short-lived device code (stored only as a hash) and a user code, which you approve while signed in at /dashboard/link. Codes expire, and code requests are rate-limited. Revoke a key at any time from settings to cut off CLI access immediately.
We use the data above only to provide and operate Procli Code: to authenticate you, connect repositories, run scans, verify agent claims, assess agent actions, produce findings, remember your preferences, meter usage, keep the service reliable and secure, prevent abuse, and support you.
We do not sell your data or share it for advertising, and we do not use your private repository contents to train models for other customers.
A small set of processors help run the product: Clerk (authentication), GitHub (repository connection through the Procli GitHub App), the Procli engine (Guard and Cheat data), and our hosting and database infrastructure. Each processes data only to provide its part of the service.
Sensitive material is kept server-side and out of the browser: GitHub installation tokens, the engine signing key, and API-key hashes are never exposed to the client. Connections to the engine are encrypted and use short-lived, per-account tokens, and mutating dashboard actions require a same-origin request and are rate-limited.
No system is risk-free. Avoid pointing scans at secrets, private environment files, or regulated data you do not want processed, and rotate any API key you believe has been exposed.
Scan results expire automatically. Account, repository-connection, key, and preference records are kept while your account is active and as needed to run the product, then deleted or de-identified within a reasonable period unless a longer period is required for security, legal, or operational reasons.
You are in control: disconnect a repository or uninstall the GitHub App to stop scanning it, revoke an API key to cut off CLI access, and contact us to request deletion of your account data.
We may update this policy as the product leaves beta and evolves. The “Last updated” date above reflects the current version, and material changes will be communicated where appropriate.
Questions about this policy, or requests about your data, can be sent to procli.code@gmail.com.