What Chat Review is for
Chat Review is the workspace-wide log of conversations anyone has had with your configured agents. While Chat is where you do new work, Chat Review is where you look back — continue a thread, audit what was said, or understand how an agent behaved in production. Think of it as observability for your agents: the harness feature that lets you debug, measure, and improve what agents actually did — not just what you intended in the Builder.
Who uses Chat Review
What you can do
- Browse all workspace agent chats — filter and search by agent, participant, or date depending on your view
- Open a past thread and read the full message history, including tool use and artifacts created
- Continue a conversation where it left off when appropriate
- Compare behavior before and after a version change by finding chats from each period
Screenshot placeholder
Chat Review detail view: full message thread on the left, metadata (agent name, dates, participant) in header. Show at least one tool call or artifact reference in the transcript.
Chat Review vs. Lab
Do not treat Lab success alone as proof the agent is production-ready. Sample Chat Review after deploy.
Using review to improve agents
When a conversation goes wrong, ask:- Context — Did the agent lack a file, record, or Focus it needed?
- Prompt — Did it have the facts but apply the wrong rule or tone?
- Tools — Did it pick the wrong Action or fail to call one?
- Version — Which agent version was active? Would rolling back or patching the prompt fix it?
Where to go next
Versions & testing
Version control and safe iteration.
What agents are — and aren't
Why observability matters for non-deterministic systems.