What happens if an agent crashes mid-task?

Its lease expires and the task returns to the queue, so another agent picks it up. Work is never lost to a crash.

Can two agents work the same repo safely?

Yes — leases prevent them taking the same task, and scope keeps them in different areas. They share the repo without clobbering each other.

TaskPeace › How do I stop my AI agents from colliding on the same files?

How do I stop my AI agents from colliding on the same files?

Give each task a lease and each agent a scope. A lease locks a task to one agent so no two pick it up; scope keeps each agent working a distinct slice of the codebase. Together they turn "agents fighting over the same files" into a fleet that divides the work cleanly.

Why collisions happen

Run several agents with no coordination and it's a team with no project plan — they grab overlapping work, edit the same files, and produce conflicting diffs you then have to untangle. The fix isn't fewer agents; it's a coordination layer.

Leases + scope, in practice

Lease: the instant an agent takes a task it's marked in-progress with a short expiry. Every other agent skips it. If the agent crashes, the lease reaps and the task is re-served — no work is stranded.
Scope: agents pull only from the part of the queue (project / area) they're assigned, so two agents never end up in the same file by accident.

TaskPeace applies both automatically — pull a task, it's leased; run several sessions, they self-divide.

Get early access → Read the fleet guide

FAQ

What happens if an agent crashes mid-task?: Its lease expires and the task returns to the queue, so another agent picks it up. Work is never lost to a crash.
Can two agents work the same repo safely?: Yes — leases prevent them taking the same task, and scope keeps them in different areas. They share the repo without clobbering each other.