Two developers, same bug, same morning

Small teamsTeam with a shared bug queue

A production error landed in the team channel. Two developers each told their agent to look into it, and both opened pull requests an hour later.

How do I stop two developers' AI agents from working on the same production bug?

With ArcRoom, the first agent calls claim_task on the work, for example 'checkout 500 on coupon apply'. The second agent's claim returns who already claimed it and when, so that developer picks the next ticket. It works when both agents describe the work in similar words. Very different wording may not match.

Also asked as: Two of us opened pull requests for the same bug. How do we avoid that? · How do I let AI agents claim a bug so no one else's agent takes it? · Can my AI agent tell that a teammate's agent is already on a ticket?

The work

The first agent calls claim_task on 'checkout 500 on coupon apply'.

ArcRoom keeps
  • Claim
The next agent gets

claim_task → already claimed by rahul (codex, 6 min ago):

Step by step

  1. The first agent calls claim_task on 'checkout 500 on coupon apply'.
  2. The second agent's claim returns 'already claimed by Rahul's agent, 6 minutes ago'.
  3. The second developer picks the next ticket.

What the agent sees

claim_task → already claimed by rahul (codex, 6 min ago):
"checkout 500 on coupon apply"

Compared with today's tools

"I'm on it" in Slack, or assign the ticket

Whoever picks up the error replies in the alert thread or assigns the ticket to themselves before starting.

Setup
0 min
Ongoing
a few seconds, if people do it before starting

ArcRoom

Every agent reads the team playbook before it works and writes to it when it finishes. Nothing to update by hand.

Setup
2 min per person (npx arcroom init)
Ongoing
None

A human claim works as well when people remember; ArcRoom's claim is made by the agent, so it also covers runs nobody announced.

Where it stops

Works when both agents describe the work in similar words. Very different wording may not match.

Try it in your project

Works with the agents your team already uses, and any agent over MCP or the HTTP API.

More in small teams