Comments
>log in to comment
Unlike typical AI logging that trusts the agent’s own output, Complytic seals a signed intent receipt before execution and a linked outcome receipt after, all verifiable with a public key. It’s an open‑source reference that lets anyone run the verifier locally without any of our infrastructure. I’d love your thoughts on the design and any missing pieces.
Open-source reference implementation: signed, hash-linked receipts for AI agent database writes. Verify it yourself.
The receipt is sealed before the write executes, signed out-of-band by a separate signer process, and verifiable by anyone holding the public key.
This is a reference implementation, not a deployed product. The code is public. You can run the verifier and the tamper tests yourself without access to any of our systems.
Enterprise security reviews now run 600+ questions including AI supplements. Policy documents do not answer them. Reviewers want artifacts.
A vendor asserting its own controls is not verification. Buyers increasingly ask for evidence they can check without trusting the vendor.
Every stalled review pulls engineers off product work for weeks, usually right before a deal closes.









