Open Cowork logo

Open Cowork

A desktop AI agent workspace with multiple models, tools, and sandboxing

Project details
Open sourceLocal-firstNo account requiredMIT

Platforms

macOS · Windows

Maker

OpenCoworkAI

Reviewed Oct 12, 2026

Alternative to

Overview

Run AI agents from a desktop app on Windows or macOS. Open Cowork connects model providers with local files, MCP tools, reusable Skills, and optional virtual machine isolation.

About

Open Cowork is a free, open source desktop workspace for running AI agents on Windows and macOS. It brings supported model providers into one app, where an agent can work with files in a chosen workspace, use MCP connectors, and run reusable Skills for tasks such as creating or editing documents and spreadsheets. The app supports Claude, OpenAI-compatible services, and several other model providers. You bring an API key for the provider you choose, and provider usage may cost extra. Open Cowork does not require its own hosted account. Open Cowork can connect to desktop tools through MCP, automate graphical interfaces, and accept remote commands through Feishu or Slack. It includes path-based workspace restrictions, with stronger virtual machine isolation available through WSL2 on Windows or Lima on macOS. Files are managed on your device, while prompts and any context sent to a model are processed by the provider you configure. Linux users can build from source, but prebuilt installers are currently offered for Windows and macOS.

Good for

  • — Running AI agents in a desktop workspace
  • — Working with local files and connected MCP tools
  • — Generating and editing documents with reusable Skills
  • — Choosing between supported model providers

Limitations

  • — You need an API key from a supported model provider, and provider usage may cost extra
  • — Prompts and context are sent to the model provider you configure
  • — Prebuilt installers are available for Windows and macOS; Linux requires building from source
  • — Virtual machine isolation requires WSL2 on Windows or Lima on macOS

Related projects