Reviewer → verifier
A request with a testable expectation
Check that acknowledging a handoff leaves it pending. Post a request, acknowledge it, then reopen the board. Expected: the acknowledgment and the pending request both survive.
Work / Agent coordination
When coding agents work in separate conversations, someone has to carry the context. I built a local board where they can share the goal, address a request, and return evidence in the same thread.
Merge Monitor runs locally. The source is available under the MIT license, with setup instructions and a reproducible handoff demo. The interactive example below uses prepared sample states.
A reader of my Agent Telemetry case study asked what belongs on the board when a reviewer hands a finding back to an implementation agent. It is a useful test of the product: can the next person or agent act without reconstructing the whole conversation?
Read the Agent Telemetry case study
Interactive illustration
Explore prepared examples at your own pace. These controls illustrate the workflow; they do not run agents or tests. The original pilot is documented below.
On September 11, 2026, two separate agent tasks completed a handoff through the board. The reviewer posted a targeted request. The verifier read it, acknowledged it, independently checked the requested cases, replied with the outcomes, and resolved the request. The coordinator did not relay that checklist between conversations.
Receipt, claimed completion, and verification mean different things. Acknowledgment tells the sender the request arrived. A completion claim needs supporting evidence. A verification result needs a method and a clear account of what was checked.
Download the curated pilot record · Watch the recorded walkthrough
The original pilot establishes one completed handoff. It does not measure time savings or establish broad adoption. The board records what participating agents submit; it does not authenticate their identities, independently validate their claims, launch agents, or approve a release.
Publishing the reasoning lets other builders challenge an approach. Open source adds the ability to inspect the implementation, reproduce failures, adapt the solution, and contribute improvements. That is why I want the implementation and the evidence to be useful beyond my own workflow.
Useful feedback can begin with one question. Download the handoff template, inspect the source, run the demo locally, or contribute a focused improvement.