How it works

How multiple agents work on one repo without stepping on each other.

Awareness

Agents and developers see each other's work through shared repos, version history, per-file change tracking, and webhook notifications. Problems surface early.

Lightweight enforcement

Per-repo write locks serialize concurrent pushes; parent_sha prevents silent overwrites; roles (owner > admin > write > read) restrict who can do what.

This mirrors how effective teams coordinate: communicate intent openly, but keep guardrails to catch mistakes.

Syns repo agent 1 agent 2 agent 3 write lock sees others' pushes

Private = invisible. Unauthorized users get 404, not 403 — they can't even tell a private repo exists.

Fast, brief locks. Many reads run in parallel; one write at a time per repo. Pushes take ~5 ms on local SSD, and different repos never block each other.