Open Cowork
Open-source AI agent desktop app for Windows & macOS. One-click install Claude Code, MCP tools, and Skills — with sandbox isolation, multi-model support, and Feishu/Slack integration.
What is Open Cowork?
Open Cowork is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to open-source ai agent desktop app for windows & macos. one-click install claude code, mcp tools, and skills — with sandbox isolation, multi-model support, and feishu/slack integration.
Open-source AI agent desktop app for Windows & macOS. One-click install Claude Code, MCP tools, and Skills — with sandbox isolation, multi-model support, and Feishu/Slack integration.
This server falls under the Developer Tools category on MCPgee, the world's largest MCP server directory with 33,000+ servers.
Features
- Open-source AI agent desktop app for Windows & macOS. One-cl
Use Cases
Maintainer
Works with
Installation
Manual Installation
npx open-coworkConfiguration
Configuration Details
claude_desktop_config.json
Performance
Response Metrics
Resource Usage
How to Set Up and Use Open Cowork
Open Cowork is an open-source desktop AI agent application for Windows and macOS built on Electron, providing a full GUI environment for running Claude and other LLMs as autonomous coding and productivity agents. It ships with built-in sandbox isolation (WSL2 on Windows, Lima VM on macOS), one-click MCP tool setup, multi-model provider support including Anthropic, OpenRouter, Kimi, and GLM, and native integration with Feishu and Slack for remote agent control. Teams and individual developers use it as a self-hosted alternative to cloud AI desktops, with document generation skills (PPTX, DOCX, XLSX, PDF), GUI automation, and browser MCP connectors all available out of the box.
Prerequisites
- macOS or Windows 10/11 desktop operating system
- An AI provider API key: Anthropic (console.anthropic.com), OpenRouter, Kimi, GLM (bigmodel.cn), or MiniMax
- Homebrew on macOS for the recommended cask install, or Node.js 18+ if building from source
- Optional: Lima (macOS) or WSL2 (Windows) for enhanced sandbox isolation
Install Open Cowork via Homebrew (macOS)
The fastest way to install on macOS is through the official Homebrew tap. The --no-quarantine flag is required because the app is not notarized through the Mac App Store.
brew tap OpenCoworkAI/tap
brew install --cask --no-quarantine open-coworkBuild from source (Windows or alternative)
On Windows or if you prefer a source build, clone the repository, install Node.js dependencies, and run the development server or build a distributable installer.
git clone https://github.com/OpenCoworkAI/open-cowork.git
cd open-cowork
npm install
npm run rebuild
npm run devConfigure your AI provider
On first launch, open Settings and enter your API key, provider base URL, and model name. For Anthropic, use the default base URL; for OpenRouter use https://openrouter.ai/api. Set the workspace folder where the agent will read and write files.
Enable sandbox isolation (recommended)
For enhanced security, install Lima on macOS so Open Cowork can run agent operations inside an isolated VM. The app manages the 'claude-sandbox' VM automatically after Lima is installed.
brew install limaAdd MCP connectors and skills
From the MCP panel in the app, enable browser automation, Notion, or custom MCP servers with a single toggle. Built-in skills for PPTX, DOCX, XLSX, and PDF generation are available immediately without extra configuration.
Connect Feishu or Slack for remote control
To trigger agent tasks remotely, configure the Feishu or Slack integration in settings by providing your bot token and channel. You can then send messages to the bot from your phone or any device to start agent workflows.
Open Cowork Examples
Client configuration
Open Cowork is a standalone desktop app and does not require a separate MCP client config file. The equivalent settings in the app UI map to these conceptual parameters.
{
"provider": "anthropic",
"apiKey": "sk-ant-your-key-here",
"model": "claude-sonnet-4-5",
"workspaceFolder": "/Users/you/projects/my-project",
"sandboxLevel": "enhanced"
}Prompts to try
Example prompts you can type into Open Cowork's agent chat to demonstrate its capabilities.
- "Read the financial_report.csv in my workspace and create a PowerPoint summary with 5 slides."
- "Organize the files in my Downloads folder by file type and move them into dated subfolders."
- "Generate an Excel spreadsheet from the data in report.json with charts."
- "Open Chrome, navigate to our staging site, take a screenshot, and save it as screenshot.png."
- "Clone the repo at https://github.com/example/app, run the tests, and summarize any failures."Troubleshooting Open Cowork
macOS Gatekeeper blocks Open Cowork from opening
Re-install using 'brew install --cask --no-quarantine open-cowork'. If already installed, run 'xattr -dr com.apple.quarantine /Applications/Open\ Cowork.app' in Terminal to remove the quarantine flag.
API key is accepted but the agent returns authentication errors at runtime
Double-check the Base URL in settings matches your provider. For OpenRouter use https://openrouter.ai/api; for Kimi use https://api.kimi.com/coding/; for Anthropic leave it as the default. Mismatched base URLs cause auth failures even with a valid key.
Lima sandbox VM fails to start on macOS
Run 'limactl start claude-sandbox' manually in Terminal to see the full error. Common causes are insufficient disk space or a conflict with Docker Desktop's VirtioFS driver. Ensure Docker Desktop is not set to use VirtioFS if Lima also needs the hypervisor.
Frequently Asked Questions about Open Cowork
What is Open Cowork?
Open Cowork is a Model Context Protocol (MCP) server that open-source ai agent desktop app for windows & macos. one-click install claude code, mcp tools, and skills — with sandbox isolation, multi-model support, and feishu/slack integration. It connects AI assistants to external tools and data sources through a standardized interface.
How do I install Open Cowork?
Follow the installation instructions on the Open Cowork GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.
Which AI clients work with Open Cowork?
Open Cowork works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.
Is Open Cowork free to use?
Yes, Open Cowork is open source and available under the MIT license. You can use it freely in both personal and commercial projects.
Open Cowork Alternatives — Similar Developer Tools Servers
Looking for alternatives to Open Cowork? Here are other popular developer tools servers you can use with Claude, Cursor, and VS Code.
Ecc
★ 188.2kThe agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
Javaguide
★ 155.8kJava 面试 & 后端通用面试指南,覆盖计算机基础、数据库、分布式、高并发、系统设计与 AI 应用开发
Gemini CLI
★ 104.5kA secure MCP server that wraps the Google Gemini CLI, allowing clients to query Gemini models using local OAuth sessions without requiring an API key. It provides tools for model interaction and diagnostics with built-in protection against command in
Awesome MCP Servers
★ 87.3k⭐ Curated list of Model Context Protocol (MCP) servers - tools that extend Claude Desktop, Cursor, Windsurf, and other MCP clients with custom capabilities.
MCP Servers
★ 86.0kModel Context Protocol Servers
CC Switch
★ 77.5kA cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Gemini CLI & Hermes Agent. Only official website: ccswitch.io
Browse More Developer Tools MCP Servers
Explore all developer tools servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.
Set Up Open Cowork in Your Editor
Choose your AI client for step-by-step setup instructions.
Quick Config Preview
Add this to your claude_desktop_config.json or .cursor/mcp.json
Ready to use Open Cowork?
Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.