open-computer-use

Overview

Open Computer Use is a comprehensive MCP server that gives LLMs access to a fully-equipped Ubuntu sandbox environment running in isolated Docker containers. It enables AI agents to execute code, create documents, browse the web with real-time streaming, and run autonomous sub-agents through Claude Code. The solution is designed for production multi-user deployments with proper container isolation between chat sessions. It integrates seamlessly with any MCP-compatible client including Open WebUI, Claude Desktop, LiteLLM, and n8n.

Try asking AI

After installing, here are 6 things you can ask your AI assistant:

• execute_command

  Execute commands in the sandbox terminal

  Try:Run 'npm install' in the project directoryTry:List all files in the current directory
• create_document

  Create documents using various document skills

  Try:Create a PowerPoint presentation about quarterly resultsTry:Generate a Word document with project requirements
• create_spreadsheet

  Create Excel spreadsheets with formulas

  Try:Build a sales tracking spreadsheet with chartsTry:Create a budget projection with formulas
• create_pdf

  Create, fill forms, extract, or merge PDFs

  Try:Convert this Word document to PDFTry:Extract text from this PDF invoice
• playwright_cli

  Browser automation and web scraping using Playwright

  Try:Scrape data from this websiteTry:Automate form submission on example.com
• describe_image

  Analyze images using vision API

  Try:Describe what's in this imageTry:Extract text from this screenshot
• frontend_design

  Build production-grade user interfaces

  Try:Create a landing page for a new coffee shopTry:Design a dashboard for analytics
• webapp_testing

  Test web applications with Playwright

  Try:Run tests on my web applicationTry:Check if login functionality works
• sub_agent

  Delegate complex tasks to Claude Code sub-agent

  Try:Create a React app for data visualizationTry:Set up a CI/CD pipeline for this project
• skill_creator

  Create custom skills for repetitive tasks

  Try:Create a skill for generating weekly reportsTry:Build a skill to format code according to standards
• gitlab_explorer

  Explore GitLab repositories

  Try:List issues in this GitLab projectTry:Show recent commits for the repository
• build_chart

  Create data visualizations from datasets

  Try:Create a bar chart showing quarterly revenueTry:Build a line graph showing user growth over time

Comparable tools

Installation

git clone https://github.com/Wide-Moat/open-computer-use.git
cd open-computer-use
cp .env.example .env
# Edit .env — set OPENAI_API_KEY (or any OpenAI-compatible provider)

# 1. Start Computer Use Server (builds workspace image on first run, ~15 min)
docker compose up --build

# 2. Start Open WebUI (in another terminal)
docker compose -f docker-compose.webui.yml up --build

FAQ

Which LLM providers are supported?

Any OpenAI-compatible provider including OpenAI, Anthropic, OpenRouter, qwen, DeepSeek, and 75+ others.

How is user data isolated between sessions?

Each chat session runs in its own isolated Docker container with no data leakage between users.

Can I customize the tools available in the sandbox?

Yes, you can customize the Docker image or create custom skills using the built-in skill creator.

Compare open-computer-use with