openTag

Your model, your machine, your AI coworker

Website · Quick Start · Docs · Contributing · Security

English | 简体中文

About

openTag is an open-source, multi-model AI coworker. From Slack and Lark talk to AI agents that run on your own machine and use your model provider of choice.

• Message an agent in Slack or Lark - it runs locally and replies in the channel
• Agents can communicate with each other, one agent can delegate to others and check progress
• Bring your own Codex, Claude Code, or Pi agents
• Open-source and self-hostable

Quick Start

You'll need a Mac or Linux computer, a Codex, Claude Code, or Pi agent, and Slack or Lark installed.

1. Create your account

Open app.opentag.build and sign in.

2. Create an agent

Go to Agents and follow the steps to choose your agent and messaging app.

3. Connect your computer

Copy the connection command shown during setup. On the computer where you want your agent to work, open the Terminal app, paste the command, and press Enter. Follow the prompts to finish connecting and sign in to your selected Runtime if asked. For Pi, install pi 0.80.6 or newer and configure a usable model provider. Keep this computer on while your agent is working.

4. Start chatting

Finish the setup steps to add openTag to Slack or Lark, then send your agent a message in a channel or direct message. For Slack, see the additional setup instructions.

Self-hosting and local development

openTag is open source and self-hostable. To run the server locally from a checkout, follow Run locally from source in the development guide.

Contributing

Issues and pull requests are welcome; start with the Contributing guide and Code of Conduct, and report vulnerabilities through the Security policy.

License

openTag is licensed under the Apache License 2.0. See the trademarks and asset provenance record for third-party marks and assets.