mcpField
W
Windows Crate Docs Search MCP

Windows Crate Docs Search MCP

Local
github.com/preludeorgMemory & Knowledge14Updated 55w ago

About Windows Crate Docs Search MCP

Windows Crate Docs Search MCP is an MCP server for searching Rust `windows` crate API documentation, aimed at developers looking up Windows APIs. It retrieves details for exact matches and lists related API items.

Key features

  1. 01Search Windows APIs
  2. 02Retrieve documentation for exact matches
  3. 03List related API items

Use cases

  1. 01Find functions in the `windows` crate
  2. 02Check type signatures and descriptions
  3. 03Identify the required Cargo feature for an API

MCP support

It provides MCP clients with the `search_windows_api` tool to search the documentation and retrieve details for exact matches.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"windows-rs": {
"command": "uvx",
"args": [
"windows-rs-mcp"
]
}
}
}
Run commanduvx windows-rs-mcp

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
pypiwindows-rs-mcp

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

Windows Crate Docs Search MCP

This project provides a Claude Desktop MCP server that allows searching the Rust windows crate API documentation hosted at microsoft.github.io/windows-docs-rs.

It uses Playwright to interact with the documentation website's search functionality and the features table.

Features

  • Search Windows API: Use the search_windows_api tool to find functions, structs, enums, etc., within the windows crate docs.
  • Get Documentation: Retrieve detailed documentation for exact matches, including description, signature, and the required Cargo feature (best effort).
  • Find Related Items: List related API items found during the search.

Usage

This package requires Playwright browsers to be installed. The server attempts to install the default browser (chromium) automatically on the first run. You can skip this by setting the MCP_SKIP_PLAYWRIGHT_INSTALL environment variable.

Running the MCP Server

Using uvx:

The simplest way to run the server after installation is using uvx:

uvx windows-rs-mcp

Running Locally with uv:

If you have cloned the repository, you can run the server directly using uv run from the project directory:

# Navigate to the project root directory
# cd /path/to/windows-rs-mcp  (Linux/macOS)
# cd C:\path\to\your\project\windows-rs-mcp (Windows)

# Run the server
uv run windows-rs-mcp
Configuring Claude Desktop

To use this MCP with Claude Desktop, add the following configuration to your Claude Desktop settings:

Option 1: Using uvx (Recommended)

{
  "mcpServers": {
    "windows-docs": {
      "command": "uvx",
      "args": [
        "windows-rs-mcp"
      ]
    }
  }
}

Option 2: Using uv run (For local source)

Make sure to replace the placeholder path in args with the actual absolute path to your project directory.

{
  "mcpServers": {
    "windows-docs": {
      "command": "uv",
      "args": [
        "--directory",
        "C:\\path\\to\\your\\project\\windows-rs-mcp", // <-- Replace with your path
        "run",
        "windows-rs-mcp"
      ]
    }
  }
}

After configuring, restart Claude Desktop. You should then be able to use the search_windows_api tool.

Configuring Cursor

Open .cursor/mcp.json and add the MCP server

{
  "mcpServers": {
    "windows-rs-mcp": {
      "command": "uvx",
      "args": ["windows-rs-mcp"]
    }
  }
}

Environment Variables

  • MCP_SKIP_PLAYWRIGHT_INSTALL: Set to true, 1, or yes to skip the automatic Playwright browser installation check.
  • MCP_PLAYWRIGHT_BROWSER: Specifies the browser Playwright should install (defaults to chromium). Other options include firefox and webkit.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.