Coordinate **one change across many repos**, or build **a new system from scratch**, with [GitHub Copilot CLI](https://github.com/github/copilot-cli) or [Claude Code](https://code.claude.com).

Website: [openfleet.tools](https://openfleet.tools) · Repository: [github.com/yashicon21/openfleet-tools](https://github.com/yashicon21/openfleet-tools)

You describe the change once. A **coordinator** agent turns it into a spec with you, then runs a **planner → implementer → tester** pipeline for each repo, monorepo package or path-scoped unit (such as one page of an app), in parallel, each in its own git worktree. A repo with several children gets its own **repo coordinator**, which merges them into one integration branch and one PR. Every agent action is recorded in an append-only event log.

No repos yet? Describe the system instead: OpenFleet proposes the services, contracts and stack, creates the repos and builds an MVP slice that one end-to-end test proves (see [Greenfield projects](#greenfield-projects)).

OpenFleet is built from the host's own primitives (custom agents, subagents, hooks, skills). It is not a new agent runtime.

Workflow at a glance

1. Grill. npx openfleet new "" starts the coordinator. It asks you questions and writes spec.md (repos and children, contracts, shared approach, baselines, branches, parallelism, completion mode). Your answer to the final "anything to change?" question starts the run. Everything after that is hands-off.
2. Plan. Every child's planner runs at once. The coordinator validates each plan against the spec.
3. Gate. All plans are checked against the spec's shared approach and against each other (per repo, then across repos). Then plans-ready extensions run.
4. Baseline (optio