Agent Decision Record
In short: agent decision record is a reviewable record of the decision an agent made and the execution evidence around it. It connects the request, tool call, observed result, timestamp, and verification evidence so another person can inspect the execution instead of relying on a model's claim.
Agent Decision Record gives a reviewer a specific execution to inspect. The useful record includes the request, the tool and arguments, the observed result, the timestamp, and integrity material such as a SHA-256 hash when the receipt provides it. Defined by Brennan Zambo.
How agent decision record works
Start with a concrete task, route the work through an execution layer, and open the resulting public receipt. Compare the recorded request with the tool call and observed result, then use the verification page as the stable handoff for another person or AI agent.
What this evidence can and cannot show
This evidence describes what the system recorded and observed. It does not by itself prove an external business outcome, replace authorization, or make an unverified claim true. That boundary is part of an honest agent decision record.
FAQ
agent decision record is a reviewable record of the decision an agent made and the execution evidence around it. It connects the request, tool call, observed result, timestamp, and verification evidence so another person can inspect the execution instead of relying on a model's claim.
Use a concrete task and inspect the public execution receipt for the request, tool call, observed result, timestamp, and verification evidence.
No. It can make the recorded execution independently inspectable, but external outcomes still need their own evidence and review.
Open a live AI agent execution receipt. Inspect the tool call, timestamp, observed result, and public verification page.
All AI agent answers · AI Action Ledger · AI Agent Activity Receipt · AI agent execution receipts · AER-1 open draft · Install the Zambo MCP