Policy for risky surfaces
Configure which paths and actions need clearance: auth, billing, infra, CI, secrets-related files, and deploys.
Tul-KI fingerprints each risky agent action, checks policy, requests approval when required, and issues a signed Action Receipt that downstream gates can verify before they accept the change.
Configure which paths and actions need clearance: auth, billing, infra, CI, secrets-related files, and deploys.
Each Action Receipt binds approver, policy, changed paths, and diff hash. Change the code and clearance expires.
Verify the same receipt in PR checks, merge rules, pilot deploy checks, or other acceptance points.
Tul-KI fingerprints the diff and reads the affected paths, workflow, and policy context.
Configured low-risk edits can pass. Teams can pause sensitive paths and actions for approval.
The signed receipt is valid only for that diff, so modified or stale changes must be cleared again.
tulki doctor maps where agents can affect sensitive code, CI, deploy, and approval flows, then gives teams a starting policy before enforcement.
Tul-KI checks the receipt against the current diff. It does not trust a PR title, branch, or stale approval; if the diff changes, clearance no longer applies.
No valid Action Receipt found for this exact diff.
We’re looking for teams already using AI coding agents on production repos. In the pilot, we’ll map your risky paths and actions, write the first policy, and wire exact-diff receipt checks into one acceptance point: PR, merge, deploy, CLI, or webhook.
We’ll use this only to follow up about the Tul-KI design partner program. Protected by Cloudflare Turnstile.