# WpfPilot MCP [Health: Active]

**Category:** 🖥️ OS Automation  
**Repository:** https://github.com/skuzadev/wpfpilot-mcp  
**GitHub Stars:** 2  
**npm Downloads (last month):** 98  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/wpfpilot-mcp

## Description
Automate Windows WPF apps from AI agents with semantic UIA selectors and FlaUI test generation.

## Tools
Capabilities this server exposes over MCP:

- **wpf_capabilities** — List supported verbs, query kinds, and wait conditions.
- **wpf_query** — Read UI state.
- **wpf_act** — Perform UI actions.
- **wpf_wait** — Wait for UI state.
- **wpf_assert** — Verify UI state.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "wpfpilot-mcp": {
    "command": "npx",
    "args": ["-y","@skuzadev/wpfpilot-mcp"]
  }
}
```

## Documentation

## What WpfPilot MCP MCP server does

WpfPilot MCP MCP server lets an MCP-compatible coding agent work with a running or launchable Windows WPF application. It uses Windows UI Automation and semantic selectors, so interactions can target controls and their accessible properties rather than fixed screen coordinates. This makes it suitable for local UI testing, workflow exploration, and diagnosing behavior in desktop applications.

The server can expose the application’s UI structure and state, including text, values, control state, bounds, patterns, child and ancestor relationships, sibling relationships, and selection. It also supports screenshots, workflow recording, and generation of xUnit tests that use FlaUI. The optional probe operates inside the target WPF process and can provide information that ordinary UI Automation does not expose, such as binding, ViewModel, command, validation, and dispatcher details.

## How it works

Communication uses MCP over stdio. An MCP client starts the launcher, which downloads the latest Windows release binary on its first run and proxies stdio to that binary. The server then discovers or attaches to a WPF process and exposes structured tools for reading and changing its UI.

The five documented core tools divide responsibilities clearly:

- `wpf_capabilities` reports available action verbs, query types, and wait conditions.
- `wpf_query` reads UI state and relationships.
- `wpf_act` performs operations such as clicking, setting values, selecting, toggling, expanding, collapsing, scrolling, and drag-and-drop.
- `wpf_wait` waits for a specified UI condition.
- `wpf_assert` checks UI state and returns structured errors when verification fails.

WpfPilot MCP MCP server is intended for local development and test automation. Mutating UI operations are audited under the user’s local application-data folder, and the process runs with the user account that starts it.

## Setup and configuration

The supported platform is Windows 10 or Windows 11, with a WPF application available to automate. Node.js 18 or newer is needed when using the recommended `npx` launcher; the release binary and npm launcher do not require the .NET SDK.

A standard MCP configuration runs `npx -y @skuzadev/wpfpilot-mcp` as a stdio server. The same configuration pattern is documented for Claude Desktop, Cursor, Codex, Claude Code, and VS Code, with client-specific configuration locations or commands. After connecting, a useful first request is to list the WpfPilot tools and attach to a running WPF application.

The optional probe requires explicit installation in the target application. The application must call `ProbeHost.Start()`, after which the agent can use the probe connection and status operations described by the project documentation.

## Tools and capabilities

WpfPilot MCP MCP server supports semantic inspection and interaction rather than coordinate-based automation. Queries can examine controls, values, text, state, patterns, hierarchy, bounds, and selection. Actions cover common WPF control operations and include drag-and-drop. Wait and assertion tools let an agent synchronize with UI changes and verify outcomes using structured results.

Beyond direct interaction, the project supports capturing screenshots, recording workflows, and producing xUnit plus FlaUI test code. The probe is useful when a visible UI symptom needs application-level context, including binding failures or command and validation information.

## Limitations and notes

The server is Windows-specific and requires a WPF application. Attachment depends on the application running in the same user session and at a compatible privilege level; an administrator-run application may require the MCP client to run elevated as well.

The probe is not enabled automatically. It must be installed in the target application and configured to start there. The server does not provide general shell, registry, or arbitrary filesystem tools through MCP. Review and trust the local server code before enabling it in an agent, particularly because it can interact with UI visible to the current user account.

_Full upstream README: https://allmcps.com/mcp/wpfpilot-mcp/readme_