GitHub radar
A personal AI agent that works your browser, terminal, and files
OpenMuse is a framework for a personal AI agent that browses the web, runs terminal commands, and edits files to actually finish tasks — not just tell you what to do.
It's an AI assistant that doesn't just reply with text — it browses sites, runs terminal commands, edits files, and can fill out a PDF itself, producing a completed copy.
When it helps
Useful if you want an agent to sort email, research things online, and fill out PDF documents or summarize a transactions CSV (spreadsheet file); skip it if you need a free or offline tool — the model and CopilotKit keys are paid.
Pros
- Finds aquarium exhibits and fills out a PDF trip permission slip on its own
- Every action can be reviewed and paused before it happens
- Runs on web, iOS, and Android from one codebase
- One-click deploy to Render is included
Cons
- Requires paid CopilotKit Intelligence and model keys (OpenAI, Anthropic, or Google)
- This is an alpha-stage framework, not a ready-made app
- The agent's terminal is isolated and can't see your computer's files
How to set it up — step by step
- 1Open your AI agent (Claude Code or Codex) inside your project folder.
- 2Send the agent the text from the block below.
- 3Give the agent your CopilotKit Intelligence key and a model key (OpenAI, Anthropic, or Google) when it asks.
- 4Let the agent deploy OpenMuse and start the server following the README.
- 5Check the result: ask the agent to show OpenMuse completing a test task in chat.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Deploy OpenMuse from https://github.com/CopilotKit/openmuse — follow the README, ask me for a CopilotKit Intelligence key and a model provider key (OpenAI, Anthropic, or Google), start it, and show me the agent completing a test task in chat.
Runs on a regular laptop with Node.js — but your own CopilotKit Intelligence key and a model provider key (OpenAI, Anthropic, or Google) are required, so it is neither free nor offline.
Open on GitHub▌ More finds