* refactor(cli): remove lifecycle-worker subprocess model, run polling in-process Replaces the per-project `ao lifecycle-worker` subprocess with an in-process polling loop managed inside `lifecycle-service`. All registered projects are now polled from the single long-lived `ao start` process. - Drops the `lifecycle-worker` command and its registration. - Rewrites `lifecycle-service` around a `Map<projectId, ActiveLoop>`, with SIGINT/SIGTERM/beforeExit graceful shutdown. - Removes PID-file coordination (PID file, log file, status, etc.) — these only existed to track subprocess state. - Per-project error isolation is preserved: the core lifecycle manager's `pollAll()` already catches per-cycle errors, and stop failures in one project can't prevent others from stopping. - Updates `start`/`spawn` callers and tests to drop the PID/logFile shape. - Adds `lifecycle-service.test.ts` covering idempotency, unknown projects, error isolation across projects, and graceful stop-all. Closes #1185 * fix(cli): address bugbot review on lifecycle-service in-process refactor - `ao spawn` / `ao batch-spawn` no longer call `ensureLifecycleWorker`. That call used to start `setInterval` polling in the one-shot spawn process, which (a) kept the CLI alive forever after the session spawned and (b) duplicated polling already running in `ao start`. Replaced with a `warnIfAONotRunning()` helper that checks `running.json` and prints a hint if the orchestrator isn't up. - `ao stop` no longer calls `stopLifecycleWorker`. That call always ran against a fresh in-memory map (stop is a separate process from start), so it always returned false and printed "Lifecycle worker not running" misleadingly. SIGTERM to the `ao start` PID already triggers the shared shutdown handler in `lifecycle-service`, which stops every loop. - Drop duplicated shutdown closure inside `registerSignalsOnce` — signal handlers now reference `stopAllLifecycleWorkers` directly. - Update tests accordingly: spawn.test.ts mocks `running-state.getRunning`, start.test.ts drops `stopLifecycleWorker` expectations. * fix(cli): drop SIGINT/SIGTERM listeners from lifecycle-service Installing listeners for those signals removes Node.js's default "exit on signal" behavior (per Node docs: "its default behavior will be removed — Node.js will no longer exit"). Since the registered listener doesn't call process.exit(), the `ao start` process would hang on SIGTERM with the setInterval timer keeping the event loop alive forever — effectively breaking `ao stop`. Default signal handling terminates the process cleanly; the OS reclaims the interval timer and dashboard child. `stopAllLifecycleWorkers` stays exported for callers that want explicit cleanup before exit. * fix(cli): project-scoped spawn warning + flush lifecycle health on exit Addresses reviewer feedback on PR #1186: - `warnIfAONotRunning` now takes a projectId and warns not only when no AO is running, but also when the running instance isn't polling the target project (e.g. `ao start A` then `ao spawn` in B left users silent about the fact that B wasn't being polled). - `running.json` now records only the project this `ao start` actually polls, not every project in config. Previously this list was a lie — `ensureLifecycleWorker` is called for the selected project only. - `ao start` installs SIGINT/SIGTERM handlers that call `stopAllLifecycleWorkers()` (flushing per-project "stopped" health state) and then `process.exit()`. Installing the handler safely requires an explicit exit because SIGINT/SIGTERM listeners remove Node's default exit behavior. * fix(cli): remove dead stopLifecycleWorker, add missing test mock Addresses bugbot comments on PR #1186: - Delete `stopLifecycleWorker` from `lifecycle-service.ts`. It was only called from `ao stop` in the old subprocess model; in the in-process model, SIGTERM to the `ao start` pid + the shutdown handler in `start.ts` covers cleanup. No production caller remains. - Add `stopAllLifecycleWorkers: vi.fn()` to `start.test.ts`'s `lifecycle-service.js` mock. Without it, `vi.mock` replaced the module and the named import resolved to undefined; the shutdown handler's try/catch would silently swallow the resulting TypeError, hiding any regression in how shutdown is wired. - Update `lifecycle-service.test.ts` to drop references to the removed export (one test removed, one repurposed as a no-op smoke test for `stopAllLifecycleWorkers` against an empty active map). |
||
|---|---|---|
| .changeset | ||
| .cursor | ||
| .github | ||
| .husky | ||
| .issue-assets | ||
| artifacts | ||
| changelog | ||
| docs | ||
| examples | ||
| openclaw-plugin | ||
| packages | ||
| scripts | ||
| skills/agent-orchestrator | ||
| tests/integration | ||
| .eslintignore | ||
| .gitignore | ||
| .gitignore-template | ||
| .gitleaks.toml | ||
| .npmrc | ||
| .prettierignore | ||
| .prettierrc | ||
| AGENTS.md | ||
| ARCHITECTURE.md | ||
| CLAUDE.md | ||
| CONTRIBUTING.md | ||
| DESIGN.md | ||
| LICENSE | ||
| README.md | ||
| SECURITY.md | ||
| SETUP.md | ||
| TROUBLESHOOTING.md | ||
| agent-orchestrator.yaml.example | ||
| eslint.config.js | ||
| package.json | ||
| pnpm-lock.yaml | ||
| pnpm-workspace.yaml | ||
| test-ao-config.yaml | ||
| test-ao-config2.yaml | ||
| tsconfig.base.json | ||
| tsconfig.node.json | ||
README.md
Agent Orchestrator — The Orchestration Layer for Parallel AI Agents
Spawn parallel AI coding agents, each in its own git worktree. Agents autonomously fix CI failures, address review comments, and open PRs — you supervise from one dashboard.
Agent Orchestrator manages fleets of AI coding agents working in parallel on your codebase. Each agent gets its own git worktree, its own branch, and its own PR. When CI fails, the agent fixes it. When reviewers leave comments, the agent addresses them. You only get pulled in when human judgment is needed.
Agent-agnostic (Claude Code, Codex, Aider) · Runtime-agnostic (tmux, Docker) · Tracker-agnostic (GitHub, Linear)
Quick Start
Prerequisites: Node.js 20+, Git 2.25+, tmux,
ghCLI. Install tmux viabrew install tmux(macOS) orsudo apt install tmux(Linux).
Install
npm install -g @aoagents/ao
Permission denied? Install from source?
If npm install -g fails with EACCES, prefix with sudo or fix your npm permissions.
To install from source (for contributors):
git clone https://github.com/ComposioHQ/agent-orchestrator.git
cd agent-orchestrator && bash scripts/setup.sh
Start
Point it at any repo — it clones, configures, and launches the dashboard in one command:
ao start https://github.com/your-org/your-repo
Or from inside an existing local repo:
cd ~/your-project && ao start
That's it. The dashboard opens at http://localhost:3000 and the orchestrator agent starts managing your project.
Add more projects
ao start ~/path/to/another-repo
How It Works
- You start —
ao startlaunches the dashboard and an orchestrator agent - Orchestrator spawns workers — each issue gets its own agent in an isolated git worktree
- Agents work autonomously — they read code, write tests, create PRs
- Reactions handle feedback — CI failures and review comments are automatically routed back to the agent
- You review and merge — you only get pulled in when human judgment is needed
The orchestrator agent uses the AO CLI internally to manage sessions. You don't need to learn or use the CLI — the dashboard and orchestrator handle everything.
Configuration
ao start auto-generates agent-orchestrator.yaml with sensible defaults. You can edit it afterwards to customize behavior:
# agent-orchestrator.yaml
# Runtime data is auto-derived under ~/.agent-orchestrator/{hash}-{projectId}/
port: 3000
defaults:
runtime: tmux
agent: claude-code
workspace: worktree
notifiers: [desktop]
projects:
my-app:
repo: owner/my-app
path: ~/my-app
defaultBranch: main
sessionPrefix: app
reactions:
ci-failed:
auto: true
action: send-to-agent
retries: 2
changes-requested:
auto: true
action: send-to-agent
escalateAfter: 30m
approved-and-green:
auto: false # flip to true for auto-merge
action: notify
CI fails → agent gets the logs and fixes it. Reviewer requests changes → agent addresses them. PR approved with green CI → you get a notification to merge.
See agent-orchestrator.yaml.example for the full reference, or run ao config-help for the complete schema.
Plugin Architecture
Seven plugin slots. Lifecycle stays in core.
| Slot | Default | Alternatives |
|---|---|---|
| Runtime | tmux | process |
| Agent | claude-code | codex, aider, cursor, opencode |
| Workspace | worktree | clone |
| Tracker | github | linear, gitlab |
| SCM | github | gitlab |
| Notifier | desktop | slack, discord, composio, webhook, openclaw |
| Terminal | iterm2 | web |
All interfaces defined in packages/core/src/types.ts. A plugin implements one interface and exports a PluginModule. That's it.
Why Agent Orchestrator?
Running one AI agent in a terminal is easy. Running 30 across different issues, branches, and PRs is a coordination problem.
Without orchestration, you manually: create branches, start agents, check if they're stuck, read CI failures, forward review comments, track which PRs are ready to merge, clean up when done.
With Agent Orchestrator, you: ao start and walk away. The system handles isolation, feedback routing, and status tracking. You review PRs and make decisions — the rest is automated.
Documentation
| Doc | What it covers |
|---|---|
| Setup Guide | Detailed installation, configuration, and troubleshooting |
| CLI Reference | All ao commands (mostly used by the orchestrator agent) |
| Examples | Config templates (GitHub, Linear, multi-project, auto-merge) |
| Development Guide | Architecture, conventions, plugin pattern |
| Contributing | How to contribute, build plugins, PR process |
Development
pnpm install && pnpm build # Install and build all packages
pnpm test # Run tests (3,288 test cases)
pnpm dev # Start web dashboard dev server
See docs/DEVELOPMENT.md for code conventions and architecture details.
Contributing
Contributions welcome. The plugin system makes it straightforward to add support for new agents, runtimes, trackers, and notification channels. Every plugin is an implementation of a TypeScript interface — see CONTRIBUTING.md and the Development Guide for the pattern.
License
MIT