Security & data protection

Your code is safe.
Here's how we prove it.

Your ideas and code are your most valuable asset. When a change is reviewed, it goes only to the AI providers you choose — encrypted in transit, never persisted by us, and never added to a training set by us. This is exactly what happens to your data, and the controls around it.

How your data flows

Where your code goes — and where it doesn't

01 · GitHub

Your PR opens

A pull request is opened in your repo. We read the diff through the GitHub App you installed, scoped to only the repos you connect.

02 · Our service

Review is orchestrated

We fetch the diff at review time and route it to the models you picked. We keep the diff only for the moments the review runs.

03 · The AIs you chose

Models review it

The diff is sent over encrypted API calls to the providers you selected. What a provider retains is governed by your account's terms with them; we add nothing to any training set.

04 · Back to you

Verdict returns

Each model's verdict comes back and is posted to your PR. We store the outcome and usage — not your source code.

What we retain: pull-request references, review verdicts, and usage metrics for billing. What we don't: your source code — it isn't persisted in our database after a review runs.
Controls

Defense at every layer

Encryption at rest

API keys and secrets are sealed with authenticated encryption before they touch disk. The master key lives only in the server environment, never in a repo.

Encryption in transit

Every connection — to your browser, to GitHub, to each AI provider — is TLS-encrypted end to end.

Tenant isolation

Each customer's keys, configuration, and billing are isolated and scoped to their own identity. One tenant can never reach another's data.

Least-privilege secrets

The GitHub App is scoped to only the repos you connect. Keys are resolved from the vault at run time, never logged, and scrubbed from any error output before it reaches a screen or a log.

Tamper-evident audit log

Every admin action and merge decision is written to a hash-chained, attributed trail — who did what, and when. Any edit, deletion, or reorder breaks the chain and is flagged.

Hardened web surface

Every response carries a strict Content-Security-Policy with per-request script nonces (no inline script can execute), HSTS, and clickjacking & MIME-sniffing protections. Public endpoints are rate-limited against abuse.

Access control & SSO

Invite-only onboarding, admin control over who can connect and approve, and enterprise SSO via OIDC — Okta, Entra, Google Workspace. SAML is on the way.

Your keys, your call

Use our pooled provider keys and pay only for usage, or bring your own keys at no markup. For the strictest environments, running entirely against your own cloud — so your code never leaves infrastructure you control — is on the roadmap.

  • Our keys, metered & encrypted
  • Bring your own provider keys
  • Your own cloud endpoints — Bedrock, Vertex, Azure (coming)
  • Revoke or rotate any key any time
Compliance & process

Ready for your security review

We'll complete your vendor security questionnaire, sign a DPA and MSA, and walk your team through our architecture. Formal certifications (SOC 2) are on our roadmap; we're happy to share our current posture and controls under NDA.

Working with an enterprise? Reach out and we'll set up a security review, share documentation, and scope a bring-your-own-cloud deployment for your environment.