* refactor(core): replace spawn rollback ladder with CleanupStack Replace the four nested try/catch + cleanupSpawnWorkspaceAndMetadata helper in _spawnInner with a single LIFO CleanupStack. Each side effect (reserved metadata, workspace, prompt files, runtime handle) pushes its undo as soon as the resource exists; on success we dismiss(), on failure we runAll(). Why: adding a new spawn step previously required extending every prior cleanup block. Easy to forget; no compiler check. The stack makes rollback structural — a new step pushes one cleanup, no risk of leaving prior resources behind. runAll() is fault-tolerant by design: a throwing cleanup never short-circuits the rest. Behavior is preserved. Adds characterization tests for the worker spawn rollback paths (none existed before — only spawnOrchestrator was covered): - workspace.create failure cleans reserved metadata - runtime.create failure destroys worktree + cleans metadata - postLaunchSetup failure destroys runtime + worktree + metadata - one cleanup throwing does not skip subsequent cleanups Refs #1603 (PR 1 of the ao spawn refactor plan). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * refactor(core): make CleanupStack runAll() terminal, symmetric with dismiss() Address review feedback on PR #1616: previously `dismiss()` → `push()` was a documented no-op but `runAll()` → `push()` would silently queue cleanups that fired on a subsequent `runAll()`. Asymmetric and surprising. Set `this.dismissed = true` at the top of `runAll()` so both terminal states (success via dismiss, failure via runAll) reject further pushes identically. Add a regression test pinning the new symmetric behavior. The "idempotent runAll" test continues to pass (early-return path now fires via the dismissed flag instead of the empty-stack short-circuit). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * refactor(core): surface CleanupStack errors and cover postCreate rollback Address review feedback on PR #1616: - Pass an onError callback to cleanupStack.runAll() in _spawnInner that logs cleanup failures via console.error. The previous /* best effort */ pattern silently swallowed errors during rollback; now the same errors are surfaced for debugging without changing behavior (cleanup errors still don't propagate, subsequent cleanups still run). - Add a characterization test for the workspace.postCreate failure path. This was the only rollback path without a test — the stack handled it correctly already, but pinning it down prevents regression. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> |
||
|---|---|---|
| .changeset | ||
| .cursor | ||
| .github | ||
| .husky | ||
| .issue-assets | ||
| artifacts | ||
| changelog | ||
| completions | ||
| docs | ||
| examples | ||
| handoff/pr-1466 | ||
| openclaw-plugin | ||
| packages | ||
| schema | ||
| scripts | ||
| skills | ||
| tests/integration | ||
| website | ||
| .eslintignore | ||
| .gitignore | ||
| .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 | ||
| 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
Zsh Completion
Generate the completion file from the installed CLI:
mkdir -p ~/.zsh/completions
ao completion zsh > ~/.zsh/completions/_ao
Then make sure the directory is on your fpath before compinit runs:
fpath=(~/.zsh/completions $fpath)
autoload -Uz compinit
compinit
For Oh My Zsh, install the same generated file into a custom plugin directory and add ao to your plugin list:
mkdir -p "${ZSH_CUSTOM:-~/.oh-my-zsh/custom}/plugins/ao"
ao completion zsh > "${ZSH_CUSTOM:-~/.oh-my-zsh/custom}/plugins/ao/_ao"
If you are contributing from a source checkout, you can also symlink the repo copy at completions/_ao.
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
$schema: https://raw.githubusercontent.com/ComposioHQ/agent-orchestrator/main/schema/config.schema.json
# 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.
Keep the $schema line so editors can autocomplete and validate against schema/config.schema.json.
See agent-orchestrator.yaml.example for the full reference, or run ao config-help for the complete schema.
Remote Access
AO keeps your Mac awake while running, so you can access the dashboard remotely (e.g., via Tailscale from your phone) without the machine going to sleep.
How it works: On macOS, AO automatically holds an idle-sleep prevention assertion using caffeinate. When AO exits, the assertion is released.
# agent-orchestrator.yaml
$schema: https://raw.githubusercontent.com/ComposioHQ/agent-orchestrator/main/schema/config.schema.json
power:
preventIdleSleep: true # Default on macOS, no-op on Linux
Set to false if you want to allow idle sleep while AO runs.
Lid-close limitation: macOS enforces lid-close sleep at the hardware level — no userspace assertion can override it. If you need remote access while traveling with the lid closed, use clamshell mode (external power + display + input device).
Plugin Architecture
Seven plugin slots. Lifecycle stays in core.
| Slot | Default | Alternatives |
|---|---|---|
| Runtime | tmux | process |
| Agent | claude-code | codex, aider, cursor, opencode, kimicode |
| 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