Pairing across two laptops
Small teamsTwo developers pairing on one feature, each with an agent
Sana and Omar paired on a checkout rewrite over a call, Sana on the frontend and Omar on the API, each with Claude Code on their own laptop. Twice their agents both started on the shared types file.
How do two developers pairing on one feature keep their AI agents from editing the same file?
With ArcRoom, both agents call claim_task before a part, so a shared file like the types file goes to only one of them. When one agent finishes its part, complete_task records the result, such as a new response shape, and the other agent gets it from find_task or recall. Choices made on the call go in with arcroom decide. Neither person sees the other's agent live yet.
Also asked as: Our two agents both started on the shared types file while we paired. How do we avoid that? · How do I pass one AI agent's API changes to my pair's agent on another laptop? · Can two AI agents on different laptops split one feature between them?
Both agents call claim_task before a part, so the shared types file goes to one of them.
- Reused answer
- Claim
- Saved answer
- Decision
- Recall
claim_task → already claimed by omar (claude, 3 min ago):
Step by step
- Both agents call claim_task before a part, so the shared types file goes to one of them.
- When Omar's agent finishes the API part, complete_task records the new response shape.
- Sana's agent gets it from find_task or recall instead of asking over the call.
- Choices made on the call go in with arcroom decide, so both agents follow them.
What the agent sees
claim_task → already claimed by omar (claude, 3 min ago):
"update CheckoutSession type in shared/types.ts"
Decision (today): totals computed server-side only.Compared with today's tools
VS Code Live Share + talking
Share one editor session and say out loud who takes which file.
- Setup
- 10 min
- Ongoing
- constant coordination on the call
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
Live Share is better for seeing each other's cursor live; ArcRoom stops the two agents from taking the same task.
Where it stops
Each person's last-run note shows their own last session first. Open now lists the other's live session, but not what their agent is thinking.
Try it in your project
Works with the agents your team already uses, and any agent over MCP or the HTTP API.