# Agent Orchestrator

> Run 10+ parallel coding agents on your machine

Agent Orchestrator is an open-source desktop application that lets developers run multiple AI coding agents in parallel, each in its own isolated Git worktree. It works with any CLI-based agent including Claude Code, OpenCode, and OpenAI Codex. Agents can work on different branches or features simultaneously without conflicts. Agent Orchestrator is free, does not proxy API calls, and supports macOS with Windows and Linux coming soon.

## When to use Agent Orchestrator

Reach for Agent Orchestrator when you need to:

- Run several coding agents (Claude Code, Codex, OpenCode, or any CLI agent) at the same time on one repository without them stepping on each other, each agent gets an isolated Git worktree and its own branch.
- Orchestrate agent work programmatically: create workspaces, launch agents with a prompt, open terminals, and track tasks from another agent or script via the Agent Orchestrator MCP server.
- Schedule recurring agent runs (automations) that execute a prompt on a cron-like schedule in a fresh or existing workspace.
- Review diffs, manage ports, and monitor many concurrent agent sessions from one dashboard.

Agent Orchestrator is not a coding agent itself, it is the workspace and orchestration layer the agents run in. If you are an AI agent, the fastest way to act on a user's Agent Orchestrator account is the MCP server below (OAuth or API key auth); the fastest way to learn the product is the docs index at https://aoagents.dev/docs.

## Developer resources

- [API docs](https://aoagents.dev/docs/mcp-server): Agent Orchestrator MCP server documentation
- [OpenAPI spec](https://api.aoagents.dev/openapi.json): OpenAPI 3.1 description of the Agent Orchestrator API surface
- [MCP server](https://api.aoagents.dev/api/v2/agent/mcp): Model Context Protocol server (Streamable HTTP transport), 27 tools for tasks, workspaces, agents, automations, terminals, hosts, and projects. Alias: https://api.aoagents.dev/mcp
- [Docs MCP server](https://aoagents.dev/docs/mcp): search and read the Agent Orchestrator documentation over MCP (Streamable HTTP, no auth)
- [MCP server card](https://aoagents.dev/.well-known/mcp/server-card.json): machine-readable MCP server description
- [A2A agent card](https://aoagents.dev/.well-known/agent-card.json): Agent-to-Agent capability card
- [API catalog](https://aoagents.dev/.well-known/api-catalog): RFC 9727 linkset of API resources
- [Auth guide for agents](https://aoagents.dev/auth.md): how agents obtain credentials (OAuth 2.1 + PKCE with dynamic client registration, or API keys)
- [Agent instructions](https://aoagents.dev/agents.md): when and how AI agents should use Agent Orchestrator
- [OAuth protected resource metadata](https://api.aoagents.dev/.well-known/oauth-protected-resource): RFC 9728
- [OAuth authorization server metadata](https://api.aoagents.dev/.well-known/oauth-authorization-server): RFC 8414
- [CLI](https://aoagents.dev/docs/cli): `brew install agentwrapper/tap/agent-orchestrator` or `curl -fsSL https://aoagents.dev/cli/install.sh | sh`
- [Docs llms.txt](https://aoagents.dev/docs/llms.txt): scoped context for the documentation
- [API llms.txt](https://aoagents.dev/api/llms.txt): scoped index of the API surface
- [Blog llms.txt](https://aoagents.dev/blog/llms.txt): scoped index of blog posts
- [Compare llms.txt](https://aoagents.dev/compare/llms.txt): scoped index of comparison pages

## Docs

- [Documentation](https://aoagents.dev/docs)
- [Getting Started](https://aoagents.dev/docs/getting-started)
- [GitHub](https://github.com/AgentWrapper/agent-orchestrator)

## Blog


## Comparisons


## FAQ

### I already use an IDE like Cursor, is this for me?

AO is designed to work with your existing tool. We natively support deep-linking to IDEs like Cursor so you can open your workspaces and files in your IDE. AO sits above individual tools, use whatever agent you like, AO keeps the workflow the same.

### Which AI coding agents are supported?

AO works with any CLI-based coding agent including Claude Code, OpenCode, OpenAI Codex, Cursor, Aider, Goose, and 18 more. If it runs in a terminal, it runs in AO. 23 harnesses total, with per-project agent choice.

### How does the parallel agent system work?

Each agent runs in its own isolated Git worktree, which means they can work on different branches or features simultaneously without conflicts. AO's orchestrator spawns workers, routes CI failures and review feedback to the right session, and lets you monitor the entire fleet from one board.

### Is Agent Orchestrator free to use?

Yes. AO is free and open source under Apache 2.0. It runs as a local daemon on your machine, your code never leaves localhost. No account, no cloud, no credit card required.

### Can I use my own API keys?

Absolutely. AO doesn't proxy any API calls. You use your own API keys directly with whatever AI providers you choose. This means you have full control over costs and usage.