Kuse Cowork offers an open-source, local-first desktop agent as an alternative to proprietary cowork applications. It targets users prioritizing privacy, model flexibility, and local execution, enabling users to leverage their own API keys or local models for enhanced control.

How It Works

The agent is built in Rust with the Tauri framework, ensuring a lightweight, cross-platform native application. Core operations and external command execution are sandboxed within Docker containers for security and isolation. It supports the Model Context Protocol (MCP) for seamless integration with various tools and services.

Quick Start & Requirements

Clone the repository, install Node.js dependencies (npm install), and run in development mode (npm run tauri dev). Production builds are available via npm run tauri build. Prerequisites include Node.js 18+, Rust, and Docker Desktop (essential for container isolation). A Discord community is available at https://discord.gg/Pp5aZjMMAC.

Highlighted Details

• Model Agnostic: Works with Anthropic, OpenAI, and local models via Ollama or LM Studio.
• BYOK (Bring Your Own Key): Supports direct use of user-provided API keys and local model endpoints.
• Container Isolation: Utilizes Docker for secure, isolated command execution, preventing host system interference.
• Extensible Skills: Features a pluggable skills system with built-in support for common document formats (docx, pdf, pptx, xlsx).

Maintenance & Community

The project has a recent release (v0.0.2, Jan 26, 2026) addressing API key handling and adding experimental Excel support. Community support and discussions are available via their Discord server.

Licensing & Compatibility

The project is distributed under the permissive MIT License, allowing for broad use, modification, and integration, including within commercial or closed-source products.

Limitations & Caveats

Full container isolation features are contingent on Docker Desktop being installed and running; otherwise, commands execute without isolation. The development environment setup requires manual configuration. The project is noted as "early," and users are advised caution when granting access to local folders.