Agent Orchestrator

The orchestration layer for parallel AI coding agents

English · 简体中文 · 日本語 · 한국어 · Español · Français · Deutsch · Português (Brasil)

An Agentic IDE that supervises parallel AI coding agents in isolated workspaces, with complete control and automatic feedback loops from CI failures, review comments, and merge conflicts.

What is Agent Orchestrator?

Agent Orchestrator is a meta-harness agent IDE for running AI coding agents in parallel. It gives agents like Claude Code, Codex, Cursor, Kimi Code, opencode, and others a shared workspace where their sessions, chat or terminal interfaces, branches, pull requests, and feedback loops can be supervised from one place.

The agents still do the coding. AO provides the harness around them: isolated workspaces, live terminal access, session state, PR awareness, and automatic loops that send CI failures, review comments, and merge conflicts back to the right agent. Instead of manually coordinating a pile of agent terminals, AO turns parallel agent work into a managed workflow.

Why Agent Orchestrator?

AI coding agents become much more useful when they can work in parallel, but parallel work gets messy quickly. Branches overlap, terminals get lost, CI failures need follow-up, review comments need replies, and merge conflicts have to reach the right worker.

Agent Orchestrator is built to keep that loop visible and manageable. It helps you:

• Start multiple agents from the same project without mixing their work
• Keep every session in a separate git worktree
• See which agents are working, waiting, finished, or blocked
• Route CI failures, review comments, and merge conflicts back to the right session
• Use different agent CLIs through one common supervisor

How it works

At a high level, Agent Orchestrator follows a simple loop:

1. Add a project you want agents to work on.
2. Start one or more sessions from the desktop app or CLI.
3. AO creates an isolated git worktree for each session.
4. AO launches either the agent's terminal UI or a structured Chat controller, according to the session's selected interface.
5. The local daemon watches session state, controller activity, pull requests, CI, and review feedback.
6. The desktop app and CLI show the current state and let you send follow-up instructions to the right session.

The result is a local control layer for agentic coding: agents still do the coding, while Agent Orchestrator keeps their workspaces, status, terminals, and feedback loops organized.

Features

The desktop app is the main control surface: projects on the left, active sessions in the center, and the selected session's Chat or terminal interface, pull request state, review runs, and browser preview in the inspector.

Parallel agent sessions Start multiple coding agents from the same project without mixing files, branches, terminals, or pull request state.
Live terminal control Open any session and attach to the worker terminal while keeping session summary, PR state, and follow-up actions in view.
Review feedback loop Run reviewer agents, inspect review status, and route requested changes back to the right worker session.
In-app browser preview Preview a session's local app beside the terminal so UI work, browser state, and agent output stay together.

Supported Agents

AO ships adapters for 25 worker agent harnesses:

claude-code · codex · aider · opencode · grok · droid · amp · auggie · autohand · agy · crush · cursor · qwen · copilot · goose · continue · devin · cline · kimi · muse · kiro · kilocode · vibe · pi · kimchi

Reviewer agents are configured separately. The current reviewer harnesses are:

aider · agy · amp · auggie · autohand · claude-code · codex · cline · continue · copilot · cursor · devin · droid · crush · goose · grok · kilocode · kiro · kimi · kimchi · opencode · pi · qwen · vibe

Reviewer panes stay interactive. AO opens a fresh reviewer process for each pass so the selected harness's current task context, permissions, and environment are applied at launch. Pi reviewers run with project/user resource discovery and built-in tools disabled; AO loads an extension from its own data directory that exposes only read-only checkout inspection plus structured GitHub review posting and ao review submit. Agy, Continue, Devin, Droid, Goose, Kimchi, Kimi, Qwen, and Vibe are experimental host-trusted reviewers: their native modes and settings do not provide OS isolation, and some reviewers receive autonomous settings so they can complete review tasks. Grok, Crush, Auggie, Cline, and Autohand are experimental user-approved reviewers: AO supplies a hardened reviewer role and leaves native permission prompts enabled rather than passing broad unattended flags.

If it runs in a terminal, it runs on Agent Orchestrator.

Use the agent interface that fits the moment: structured Chat or its native terminal UI.

Install

Download the latest desktop build for your platform:

After installing, open Agent Orchestrator and point it at the repository you want AO to manage. The desktop app runs the daemon for you, so no CLI is required. Installed desktop builds check for updates on launch and periodically while the app is running. See the installation guide for agent CLI setup and troubleshooting.

npm install -g @aoagents/ao
ao start

Witness AO's Journey on X

Documentation

Telemetry

Agent Orchestrator's Electron renderer sends anonymous usage events to PostHog for reliability and product understanding. PostHog session recording is disabled by default; if a time-boxed investigation enables it, local paths and local URLs are redacted before transmission. Set VITE_AO_POSTHOG_KEY to an empty string before building to disable transmission. See docs/telemetry.md.

License

Apache License 2.0. See LICENSE.