dsh-windows-mcp

dsh-windows-mcp is a plugin that lets the DeepSeek desktop agent see the screen and control a Windows PC, for people using DSH Desktop.

What it is good for

  • Letting an AI agent click and type in Windows apps
  • Having an AI take screenshots and read what is on screen
  • Automating repetitive tasks on a Windows desktop
  • Controlling when an AI is allowed to use your computer

At a glance

Who it is for
Setup difficulty
Moderate ยท Install a desktop app or browser extension

Level 3 of 5. Download and install it like any other app or extension.

Cost
Free and open source
Type
Skill or plugin
Tasks
Works with
DeepSeek Harness Desktop, Windows-MCP, uv, Windows 10/11

How to get started with dsh-windows-mcp

  1. Open DSH Desktop and go to Settings, then Plugins.
  2. Add the bundle using the package name dsh-windows-mcp or the GitHub repository.
  3. Restart DSH so the plugin loads.
  4. Open Settings, then Desktop, and use the Dependency card to download and install Windows-MCP.
  5. Restart DSH again, then switch on the Computer Use pill in the composer.

From the project's own documentation. If a step does not match what you see, the official pages have the latest instructions.

Before you try it

  • Requires Windows 10 or 11 and DSH Desktop 0.2.0-rc.2 or newer.
  • The uv tool must be on the PATH for the dependency installer to work.
  • A model that accepts image input is needed for the agent to see screenshots.
  • Computer Use starts off, and the agent can control your desktop once you switch it on.

Momentum

Rank this week
#17
Best rank
#17 week 46, 2026
Stars gained last week
+1,098
GitHub stars
1,407

GitHub numbers as of Nov 16, 2026. Updated weekly.

0 758 1,517 Nov 6, 2026 Nov 16, 2026
GitHub stars for dsh-windows-mcp: 37 on Nov 6, 2026, 1,407 on Nov 16, 2026. Source: GitHub.

Ranking history

First ranked in week 46, 2026.

Questions

Is dsh-windows-mcp free?
Yes. It is free and open source (MIT license).
Do I need to know how to code to use dsh-windows-mcp?
No. Download and install it like any other app or extension.
What does dsh-windows-mcp work with?
According to its own documentation, it works with DeepSeek Harness Desktop, Windows-MCP, uv, and Windows 10/11.
How do I get started with dsh-windows-mcp?
1. Open DSH Desktop and go to Settings, then Plugins. 2. Add the bundle using the package name dsh-windows-mcp or the GitHub repository. 3. Restart DSH so the plugin loads. 4. Open Settings, then Desktop, and use the Dependency card to download and install Windows-MCP. 5. Restart DSH again, then switch on the Computer Use pill in the composer.
What should I know before trying dsh-windows-mcp?
Requires Windows 10 or 11 and DSH Desktop 0.2.0-rc.2 or newer. The uv tool must be on the PATH for the dependency installer to work. A model that accepts image input is needed for the agent to see screenshots. Computer Use starts off, and the agent can control your desktop once you switch it on.
What is dsh-windows-mcp good for?
Letting an AI agent click and type in Windows apps, having an AI take screenshots and read what is on screen, automating repetitive tasks on a Windows desktop, and controlling when an AI is allowed to use your computer.

Similar agents to dsh-windows-mcp

  • Windows-MCP

    This is the upstream Windows control server the plugin builds on, usable with any MCP-capable AI app.

  • open-computer-use

    Works on macOS, Linux, and Windows as a general connector, not tied to one desktop app.

  • axuse

    Reads page structure instead of screenshots and covers browser as well as desktop.

  • DesktopCommanderMCP

    Focuses on terminal commands and file management for Claude rather than mouse and screen control.

  • elizaOS

    elizaOS is an open source TypeScript framework and app for building and running autonomous AI agents for developers and personal use.

  • lcu

    lcu is a plugin that lets coding agents like Claude Code and Codex CLI control desktop apps and Chrome without needing the Codex app.

Last updated Nov 21, 2026. Tracked since Nov 6, 2026. Spotted an error? Tell us.