Windows-MCP is an MCP server that lets AI agents operate Windows computers through actions such as navigating files, opening applications, controlling windows, and interacting with user interfaces. It is used for desktop automation, software testing, and other tasks that require an agent to use Windows.
Borrowing it
Nothing to install: this file belongs to CursorTouch/Windows-MCP. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/CursorTouch/Windows-MCP/main/CLAUDE.mdgit clone --depth 1 https://github.com/CursorTouch/Windows-MCPWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/cursortouch/windows-mcp/claude-md)<a href="https://agentmods.dev/instructions/cursortouch/windows-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/cursortouch/windows-mcp/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/cursortouch/windows-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/cursortouch/windows-mcp/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.01975 | $0.01975 |
| Opus 5.5 | $0.00790 | $0.00790 |
| Sonnet 5.5 | $0.00395 | $0.00395 |
| Haiku 4.5 | $0.00198 | $0.00198 |
Grade A, and why
Windows-MCP CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 89 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Project Overview
Windows-MCP is a Python MCP (Model Context Protocol) server that bridges AI LLM agents with the Windows OS, enabling direct desktop automation. It exposes 21 tools via FastMCP:
| Group | Tools |
|---|---|
| Capture | Screenshot, Snapshot, Scrape, DisplayInventory |
| Input | Click, Type, Scroll, Move (also drag-and-drop via drag=True), Shortcut, MultiSelect, MultiEdit |
| Timing | Wait, WaitFor |
| System | App, PowerShell, FileSystem, Registry, Process, Clipboard, Notification, ControlStatus |
Tool names are defined by the name= argument of each @mcp.tool(...) in src/windows_mcp/tools/; that directory is the source of truth. Note the shell tool is registered as PowerShell, not Shell. Any subset can be removed at startup with serve --exclude-tools (e.g. windows-mcp serve --exclude-tools PowerShell,Registry).
Build & Development Commands
uv sync --extra dev # Install runtime and development dependencies
uv run windows-mcp serve # Run the MCP server
uv run --extra dev ruff format . # Format code
uv run --extra dev ruff check . # Lint code
uv run --extra dev ruff check --fix . # Lint and auto-fix
uv run --extra dev pytest # Run all tests
uv run --extra dev pytest tests/test_desktop_views.py # Run a single test file
Package manager: UV (not pip). Python: 3.14+. Build backend: setuptools.
Architecture
The codebase follows a layered service architecture under src/windows_mcp/:
Entry point — __main__.py: Builds the FastMCP server, parses CLI flags, and selects the transport. Tool registration is delegated to tools.register_all(); an async lifespan initializes the Desktop, WatchDog, and Analytics singletons, which tools resolve lazily through the get_desktop / get_analytics callables.
Tools layer — tools/: One module per tool group, each exposing register(mcp, *, get_desktop, get_analytics). tools/__init__.py holds the module list and register_all(). Tool functions are thin — they normalize arguments and delegate to a service package. The @with_analytics decorator wraps each one for telemetry, making it the existing precedent for cross-cutting concerns at the tool boundary.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago Changed · +45 tokens per session 9d64b22d2968
- 2mo ago First seen · 89 lines · 1,930 tokens per session scan A 0e4bb834de42
Windows-MCP CLAUDE.md is an instructions file published in the GitHub repository CursorTouch/Windows-MCP (10,651 stars, last pushed 2d ago), licensed MIT. It adds 1,975 tokens to every session, about $0.0079 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
deepseek-harness AGENTS.md
AGENTS.md instructions for deepseek-ai/deepseek-harness, covering agents.md, pre-stable apis and released session data, repository layout, commands and host sandbox failures.