Desktop Touch Mcp
Run a Windows-only computer-use MCP so your agent can screenshot, inspect UIA, control Chrome CDP, click, type, scroll, and run VBA macros on the desktop.
Overview
Desktop Touch MCP is a MCP server for the build phase that provides 29 Windows computer-use tools including screenshot, UIA, Chrome CDP, input automation, and VBA.
What is this MCP server?
- 29 MCP tools for Windows 10/11 computer use in one npm package
- Screenshot capture, UI Automation (UIA) tree access, and Chrome DevTools Protocol control
- Keyboard, mouse click, scroll, and typing automation on native windows
- Optional VBA execution for Office-style automation paths
- Env toggles: DESKTOP_TOUCH_AUTO_GUARD, perception resources, WinEvent sidecar dormancy, JSONL diagnostic logging
- 29 tools documented in registry description
- Registry version 1.9.2
- Transport: stdio npm @harusame64/desktop-touch-mcp
Community signal: 1 GitHub stars.
What problem does it solve?
Windows-only workflows trap agents in repo edits when the real product only exposes a GUI, Excel, or Chrome surface.
Who is it for?
Solo builders on Windows 10/11 automating QA, RPA-style tasks, or Chrome-plus-desktop hybrids from Claude Code or Cursor.
Skip if: macOS or Linux agents, unattended production bots without human oversight, or teams uncomfortable with desktop security risk.
What do I get? / Deliverables
After npm MCP registration on Windows, your agent can perceive and act on live desktop UI through guarded screenshot, UIA, CDP, and input tools.
- Agent-driven screenshots and structured UI trees on Windows
- Chrome CDP and desktop input automation via MCP tools
- Optional VBA and perception sidecar telemetry when enabled
Recommended MCP Servers
Journey fit
A 29-tool Windows computer-use server primarily extends what agents can do while building and automating local dev environments. Agent-tooling is the canonical shelf for MCP packs that give models hands on keyboard, mouse, and desktop UI beyond repo edits.
How it compares
Full Windows computer-use MCP with 29 tools, not a single-purpose screenshot skill or cloud browser farm.
Common Questions / FAQ
Who is Desktop Touch MCP for?
Windows developers and agent authors who need UIA, Chrome CDP, and desktop input tools in one stdio MCP server.
When should I use Desktop Touch MCP?
Use it during build and ship when you must drive native Windows UI or Chrome for integration tests, demos, or semi-automated ops tasks.
How do I add Desktop Touch MCP to my agent?
Install @harusame64/desktop-touch-mcp via npm, add the stdio server to your MCP client on Windows 10/11, tune DESKTOP_TOUCH_* env vars, then restart the agent host.