Solutions · Monorepos
One repo, many writers
Monorepos concentrate blast radius. Shared packages and hot directories pull every agent into the same paths — coordination is how parallel spend stays productive.

Why monorepos collide harder
Package graphs create magnets. Auth, billing, shared UI — every agent “helps” there first. Without leases, overnight fleets thrash the same directories and leave reviewers with overlapping PRs that should never have been opened in parallel. Merge queues serialize the symptom; they do not stop the wasted agent work upstream.
Lease at the boundaries your agents already understand: packages, apps, or directory trees. Exclude generated output when safe. Keep the territory map visible so humans see contention before it becomes a merge-queue tax.
Hotspot lease split
Hotspots and package boundaries
Hot directories should not be one everlasting exclusive claim. Split them into leaf leases with shorter TTLs so idle agents can take adjacent work instead of queueing on the whole package. The allocator already models hotspot overlap carefully; humans see the same pressure on the map.
Wrap agent steps with managent run. Scope prediction (optional, advisory) can suggest likely paths from task text — it never silently expands a lease. Verify landings with managent enforce verify-pr when sibling branches still might collide after leases.
Map, CLI, verify
managent status / managent watch answer “who owns packages/core right now?” Operators force-release orphaned claims when a runner dies. CI and laptop sessions share the same protocol so a PR job cannot quietly overwrite a founder’s overnight agent.
Optional semantic checks help when shared packages expose contracts many callers depend on — still advisory, still honest about limits. Try /demo/semantic after the lease loop is wired.
Next reads
Monorepo guide · Territory map · Live demo · Docs
Running agents on a large monorepo? Join the waitlist.