Open Computer Use

v1.0.0Browser Automationstable

MCP server that gives any LLM its own computer — managed Docker workspaces with live browser, terminal, code execution, document skills, and autonomous sub-agents. Self-hosted, open-source, pluggable into any model.

agentsclaude-codecomputer-usellmmcp-server
Share:
80
Stars
0
Downloads
0
Weekly
0/5

What is Open Computer Use?

Open Computer Use is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to mcp server that gives any llm its own computer — managed docker workspaces with live browser, terminal, code execution, document skills, and autonomous sub-agents. self-hosted, open-source, pluggable ...

MCP server that gives any LLM its own computer — managed Docker workspaces with live browser, terminal, code execution, document skills, and autonomous sub-agents. Self-hosted, open-source, pluggable into any model.

This server falls under the Browser Automation category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • MCP server that gives any LLM its own computer — managed Doc

Use Cases

Managed Docker workspaces for agents
Live browser and terminal access
Autonomous sub-agent execution
Wide-Moat

Maintainer

LicenseNOASSERTION
Languagepython
Versionv1.0.0
UpdatedMay 21, 2026
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

Manual Installation

npx open-computer-use

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use Open Computer Use

Open Computer Use is a self-hosted, open-source MCP server that gives any LLM a fully-equipped Ubuntu sandbox inside an isolated Docker container, including a live Playwright browser with real-time CDP streaming, a terminal running tmux, code execution in Python/Node.js/Java, 13 built-in skills for document creation (Word, Excel, PowerPoint, PDF), and autonomous Claude Code sub-agents. It speaks standard MCP over Streamable HTTP on port 8081 and works with any MCP-compatible client — Open WebUI, Claude Desktop, n8n, or LiteLLM — with each chat session getting its own isolated container so nothing leaks between users.

Prerequisites

  • Docker and Docker Compose installed (Docker Desktop on Mac/Windows or Docker Engine on Linux)
  • An OpenAI-compatible API key (OpenAI, Anthropic, OpenRouter, or any compatible provider) for the OPENAI_API_KEY env var
  • At least 8 GB RAM and 20 GB free disk space for the workspace image (first build takes ~15 minutes)
  • An MCP-compatible client or the included Open WebUI
1

Clone the repository and configure environment

Clone the repository and copy the example environment file. Set at minimum OPENAI_API_KEY (or a compatible provider key) in the .env file. Optionally set OPENAI_API_BASE_URL for OpenRouter or other providers.

git clone https://github.com/Wide-Moat/open-computer-use.git
cd open-computer-use
cp .env.example .env
# Edit .env and set OPENAI_API_KEY=your-key
2

Build and start the Computer Use server

The first build downloads and compiles the Ubuntu workspace image with all tools (Python 3.12, Node 22, LibreOffice, Playwright, Claude Code CLI, FFmpeg, etc.). Subsequent starts are fast.

docker compose up --build
3

Start Open WebUI (optional, recommended for first-time use)

In a second terminal, start the bundled Open WebUI frontend. It connects to the Computer Use server automatically on localhost:8081.

docker compose -f docker-compose.webui.yml up --build
4

Configure model settings in Open WebUI

After adding a model in Open WebUI, go to Model Settings and set Function Calling to 'Native' and Stream Chat Response to 'On'. Without Native function calling, Computer Use tools will not be invoked.

5

Connect a custom MCP client (alternative)

Point any MCP client at the self-hosted HTTP endpoint. The server accepts standard MCP Streamable HTTP requests at http://localhost:8081/mcp with an X-Chat-Id header per session.

# Verify the server is up
curl -X POST http://localhost:8081/mcp \
  -H "Content-Type: application/json" \
  -H "X-Chat-Id: test" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}'

Open Computer Use Examples

Client configuration

Environment variables for the self-hosted Computer Use server. These go in the .env file in the project root.

{
  "mcpServers": {
    "open-computer-use": {
      "url": "http://localhost:8081/mcp",
      "headers": {
        "X-Chat-Id": "my-session-id"
      }
    }
  }
}

Prompts to try

Example prompts that demonstrate the sandbox's capabilities across browsing, code execution, and document skills.

- "Scrape the top 10 GitHub Trending repositories today and save the results to an Excel file."
- "Write a Python script that generates a bar chart of monthly sales data and save it as a PNG."
- "Create a professional PowerPoint presentation about renewable energy with 5 slides."
- "Open a browser, navigate to Wikipedia, and extract the introduction section of the 'Machine learning' article."
- "Run a Node.js script that fetches the current BTC price from a public API and prints it."

Troubleshooting Open Computer Use

First docker compose up --build takes too long or fails with disk space errors

The workspace image is large (~3-4 GB). Ensure you have at least 20 GB free on the Docker volume. On Docker Desktop, increase the disk image size in Settings > Resources. If the build times out, re-run the command — Docker layer caching means it resumes from where it left off.

Computer Use tools are not called even though the model responds normally

In Open WebUI Model Settings, Function Calling must be set to 'Native' (not 'Default' or 'Off'). This is the most common setup issue. Without native function calling, the model generates text describing what it would do instead of actually calling the tools.

The browser inside the sandbox is not visible or CDP streaming does not appear

The live browser viewer is served by the Computer Use server. Check that the PUBLIC_BASE_URL env var is set correctly in .env to the browser-reachable address of the server (default: http://computer-use-server:8081). If using Docker Desktop on Mac/Windows, use http://localhost:8081 instead.

Frequently Asked Questions about Open Computer Use

What is Open Computer Use?

Open Computer Use is a Model Context Protocol (MCP) server that mcp server that gives any llm its own computer — managed docker workspaces with live browser, terminal, code execution, document skills, and autonomous sub-agents. self-hosted, open-source, pluggable into any model. It connects AI assistants to external tools and data sources through a standardized interface.

How do I install Open Computer Use?

Follow the installation instructions on the Open Computer Use GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with Open Computer Use?

Open Computer Use works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is Open Computer Use free to use?

Yes, Open Computer Use is open source and available under the NOASSERTION license. You can use it freely in both personal and commercial projects.

Browse More Browser Automation MCP Servers

Explore all browser automation servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "open-computer-use": { "command": "npx", "args": ["-y", "open-computer-use"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use Open Computer Use?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides