Solutions · Polyrepos

Many repos, one control plane

Polyrepo orgs still run fleets that touch shared contracts, generated clients, and mirrored packages. Ownership must be visible across checkout boundaries.

Polyrepo does not mean no collisions

Splitting into many repos reduces some merge conflicts — and relocates others into API drift, duplicate PRs across service boundaries, and agents that “fix” the same contract in two places at once. The coordination problem moves; it does not vanish.

Managent treats each repo as territory under the same org control plane. Agents claim work where they check out; humans see who holds what across the fleet. Cross-repo intent becomes legible instead of tribal knowledge.

Org plane across checkouts

Org-scoped API keys and repo records let laptop, CI, and cloud agents in web/, api/, and infra/ speak the same lease protocol. managent run acquires in the checkout it wraps; managent status / managent watch surface ownership per repo without pretending every path lives in one tree.

Territory leases remain repository-scoped. Cross-repo semantic federation (contract identities → dependency map) is supported where configured; low-confidence inferred edges never independently hard-block. That limit is intentional — see semantic demo for the interactive shape.

Verify per repo, operate as a fleet

Require managent enforce verify-pr (or the GitHub App check) on each service’s land path so local bypasses cannot ship. Ops console tools — force-release, kill-agent, transfer — stay org-level when a runner dies mid-heartbeat in any checkout.

Start with /demo, read fleet coordination, and compare the monorepo hotspot story on /solutions/monorepos when some packages still live in a single tree.

Related

Territory map · Fleet coordination · Docs · Security