Capture the decisions made in agent chats.
Product rules are now decided in conversations with coding agents. AgentWatch, in TestChimp Studio, watches those sessions across your developers, filters for real decisions, and proposes story and scenario updates for you to approve.
14 days, no credit card. Runs in TestChimp Studio.
Problem
The hallway conversation moved into the agent chat.
“Guests must enter an address” used to be decided over coffee and forgotten. Now it is typed, almost word for word, into Claude Code or Cursor, and forgotten a different way: in one developer’s session history. The plans drift while the agents ship.
Your QA surface, kept current by the work itself.
Watches agent chats
Where decisions are made now.
Indexes local coding-agent sessions for the projects you map. Developers keep working; nothing is pasted anywhere.
Claude Code · Cursor · Codex · 20+
Filters for decisions
Signal, not chatter.
Only chunks that decide how the product should behave go further. Implementation detail is skipped.
Semantic neighbours + Jev
Proposes plan updates
You approve, it lands in git.
Decisions become one pending plan of story and scenario updates, executed into your plans folder when you approve.
One pending plan per project
A quiet loop for developers.
Watches the agents your team already uses
AgentWatch indexes sessions from Cursor, Claude Code, Codex, Copilot, Gemini, Windsurf and many more, for the projects whose folders you have mapped. On your sync cadence, it looks for what changed since the last checkpoint.
Signal, not chatter
Agent chats are long and mostly implementation detail. Each chunk is matched to its nearest stories and scenarios in your project’s shared embedding space and scored by Jev, a decision model, before any plan is written.
One plan, yours to approve
Decisions are batched into one pending plan per project. Approve it and your plans folder updates; raise a pull request and the rest of the team, and every QA workflow downstream, works from the new surface.
- Choose which agents to watch.
- Allow adds, updates or both, for stories and scenarios.
- Explore sessions and extractions in Studio.
How specs usually keep up.
| Specs updated by hand | Agent memory files | AgentWatch | |
|---|---|---|---|
| Captures decisions across all developers | Partly | No | Yes |
| Updates the shared stories and scenarios | Yes | No | Yes |
| Separates decisions from implementation chatter | Yes | No | Yes |
| Reviewed before it lands | Yes | Partly | Yes |
| Happens without anyone remembering to do it | No | Partly | Yes |
Questions about AgentWatch.
Which coding agents does it watch?
Where does it run?
Does it change plans on its own?
Do I need existing plans first?
How does it relate to MeetBots?
Your agents make the decisions. Your plans keep up.
Enable AgentWatch in Studio, map your folders, and keep shipping.
14 days, no credit card.