black box
when an ai agent goes wrong, can you prove what it actually did?
usually no
the logs are scattered everywhere
the agent's own trace is literally the thing under review
and anyone with access can quietly edit the record
so black box
- sits at a gateway outside the agent
- records every model call and tool call
- seals every entry in a sha-256 hash chain
- change one entry and the whole chain breaks
- anyone can verify the log offline
built for teams shipping agents, and the insurers covering them
-dk