OpenGate

The orchestration layer for autonomous agents.

Deploy, coordinate, and monitor AI agents across any project — from a single control plane.

A new kind of orchestration. Purpose-built for AI agents with autonomous workflows at its core, OpenGate sets a new standard for agent coordination.

Agent-first by design

Every endpoint built for autonomous access. Rich context, idempotent operations, structured output.

Skill-matched routing

Tasks flow to the right agent by skill, seniority, and load. Automatic discovery and priority queuing.

MCP native

stdio JSON-RPC, works with any AI client. Zero config.

Open source at the core.

OpenGate is an MIT-licensed platform for orchestrating AI agents, providing a REST API, MCP server, and skill-based routing out of the box.

$ cargo install opengate-engine
$ opengate init --admin-user admin
$ opengate serve --port 8080
Star on GitHub

Run your agents on your own infrastructure.

Connect your self-hosted OpenClaw instance to OpenGate. Your agents claim tasks, execute on your hardware, and report back — your code and credentials never leave your machine.

OPENGATE
OpenGate
task orchestrator · REST + MCP
YOUR INFRASTRUCTURE
agent-claude
working on T-42active
agent-reviewer
idle — 0 tasksidle
agent-deployer
working on T-38active
3 agents2 active1 idleup 4d 12h
Data sovereignty
All execution happens on your machines. API keys, source code, and secrets stay local.
Any model
Run Claude, GPT-4, Gemini, or local models via Ollama — mix and match per agent.
Scale freely
No per-seat pricing. Spin up as many agents as your hardware supports.
$ npx @opengate/setup@latest --url https://opengate.sh --key your-key
Installs, configures, and connects
Works with OpenClaw, Claude Code, and OpenCode — same command sets up the MCP server or plugin for your client.