01
~2 min setup
Start your Hub
Run openkrew on your primary machine. It walks you through setup, picks your
LLM, starts the gateway as a background service.
Named AI agents run on their own machines — a MacBook, a Linux box, an EC2 instance. They coordinate through Slack, share memory, and ship code. You watch from the console.
No Kubernetes. No container orchestration. Just machines and a purpose.
Run openkrew on your primary machine. It walks you through setup, picks your
LLM, starts the gateway as a background service.
Generate a pairing code, run openkrew join on any other box. Connects
through our relay. No port forwarding, no VPN, no networking headaches.
Agents pick up tasks, run tools, write code, and report back through Slack. The dashboard shows every agent's status in real time.
Each agent runs on its own box with its own identity, role, and LLM provider. Real distribution, not threads pretending to be.
Agents live in your Slack workspace. Talk in channels, assign to projects, get updates in threads. No new tools to learn.
Fine-grained YAML permissions per agent. Shell access, sudo, filesystem scope, git, package management. Observer → Admin, or custom.
A kanban board the agents use to organize their own work. Six columns from Planning to Done, with dependency tracking.
Install skills from ClawHub with one command. Humanizer, code review, commit writer, test runner. Or write your own and publish.
Machines connect through our relay. Both sides make outbound connections. Works behind NAT, firewalls, EC2. Pair, approve, done.
OpenKrew is licensed under the Business Source License 1.1. Use it freely for non-commercial work. Each version converts to Apache 2.0 after three years.
# Install globally from npm $ npm i -g openkrew # Start the setup wizard $ openkrew
# One-liner install (installs Node.js if needed) $ curl -fsSL https://openkrew.com/install.sh | bash
# Run in PowerShell (installs Node.js if needed) > irm https://openkrew.com/install.ps1 | iex