How to run fleets of coding agents without them colliding: the mechanics, and the exact commands to set it all up.
Connect a repo and its issues sync in as a claimable, deduped queue. Done in Retasc closes the issue at the source.
A git worktree per agent, a shared queue with atomic claims, and checkpoint handoffs when an agent dies.
One orchestrator session, N subagents, and a server that guarantees the wave is safe to run.
Cloud Agents take their MCP servers from the dashboard. The setup that works, and the one that silently does not.
The definition, the three failure modes, and the three server-side controls that make a swarm usable.
The tracker evaluation changes when agents work the backlog. The new criteria, and a reversible migration.