Orchestrate Parallel AI Coding Agents Spawn multiple AI agents working simultaneously on different tasks across your codebase. Each agent gets its own isolated workspace, handles CI failures automatically, and pushes notifications only when human judgment is needed.
<span className="text-[#848c84]">ao spawn</span> my-project <span className="text-blue-400">123</span></div> <div className="text-gray-500 dark:text-gray-500">✓ Created worktree at ~/.worktrees/my-project-1</div> <div className="text-gray-500 dark:text-gray-500">✓ Started tmux session: my-project-1</div> <div className="text-gray-500 dark:text-gray-500">✓ Launched Claude Code agent</div> <div className="text-green-400 mt-4">Session spawned: my-project-1</div> <div className="text-gray-400 dark:text-gray-400 mt-4"> ao status
my-project-1 working feat/issue-123
my-project-2 pr_open feat/issue-124
Quick Start Get Agent Orchestrator running in minutes
Install Agent Orchestrator
Clone the repository and run the setup script to install dependencies and configure your environment. git clone https://github.com/ComposioHQ/agent-orchestrator.git cd agent-orchestrator bash scripts/setup.sh
Requires Node.js 20+, Git 2.25+, tmux, and the gh CLI for GitHub integration.
Initialize your project
Run the interactive init command to create your configuration file, or start from a repository URL. # Option A: From existing local repo cd ~/your-project ao init --auto # Option B: From a GitHub URL ao start https://github.com/your-org/your-repo
This creates agent-orchestrator.yaml with auto-detected settings for your project.
Start the orchestrator and spawn an agent
Launch the web dashboard and spawn your first agent session. ao start ao spawn my-project 123
The dashboard opens at http://localhost:3000 where you can monitor all active sessions in real-time. Explore by Topic Deep dive into Agent Orchestrator’s capabilities
Core Concepts Understand the plugin architecture, session lifecycle, and how the orchestrator coordinates parallel agents.
CLI Reference Complete command reference for spawning agents, managing sessions, and controlling the orchestrator.
Plugin System Explore the 8 plugin slots: Runtime, Agent, Workspace, Tracker, SCM, Notifier, Terminal, and Lifecycle.
Configuration Guide Configure projects, reactions, notifications, and per-project overrides in agent-orchestrator.yaml.
Key Features Everything you need to orchestrate AI coding agents at scale
Plugin Architecture Eight swappable plugin slots for runtime, agent, workspace, tracker, SCM, notifier, terminal, and lifecycle management.
Parallel Execution Run multiple agents simultaneously, each in isolated git worktrees with separate branches and PRs.
Auto Reactions Automatically handle CI failures and review comments. Agents fix issues autonomously and notify you only when needed.
Agent Agnostic Works with Claude Code, Codex, Aider, OpenCode, and any AI coding tool through the Agent plugin interface.
Resources Learn more and get support
Example Configs Ready-to-use configuration examples for GitHub, Linear, multi-project setups, and auto-merge workflows.
Troubleshooting Solutions to common issues with tmux sessions, git worktrees, and agent connectivity.
Contributing Learn how to build custom plugins and contribute to the Agent Orchestrator project.
Ready to orchestrate your agents? Start managing parallel AI coding agents across your codebase today.
Get Started Now 