hypothesi/mcp-server-tauri

MCP Server Tauri – AI‑augmented development for Tauri apps

What it is – A Model Context Protocol (MCP) server that talks to AI assistants (Claude, Cursor, Windsurf, etc.) and gives them live insight into a running Tauri v2 application. It streams screenshots, DOM snapshots, console logs and Tauri IPC traffic, and exposes a set of commands the assistant can invoke to automate the UI, inspect or emit IPC messages, and even list mobile simulators.

Why it matters – Modern AI coding assistants need concrete, up‑to‑date context to reason about a GUI app. By feeding the assistant visual and structural data from the Tauri webview and the native side, the assistant can:

• Diagnose UI bugs automatically
• Generate or fix code that interacts with the app’s IPC layer
• Drive the UI (click, type, scroll) without external tools like Selenium
• Work on desktop, Android, and iOS from a single bridge

Key capabilities

Category	Tools (21 total)
UI Automation	Screenshots, element finder, click/scroll/typing, visual picker, DOM snapshot, CSS style query, JS execution, window management
IPC Monitoring	Real‑time capture of Tauri IPC calls, state inspection, custom event emission
Mobile Dev	List iOS simulators and Android emulators
Logs	Stream console, Android logcat, iOS logs, system logs
Setup	/setup command that adds the Rust bridge plugin to a Tauri project

How you use it

1. Install prerequisites – Node 20+, Rust/Cargo, Tauri CLI, and (optionally) Xcode/Android SDK.
2. Add the MCP to your AI assistant with the aix CLI:

   npx -y @a1st/aix add mcp tauri --command 'npx @hypothesi/tauri-mcp-server' --user
   

   The assistant then knows how to start the server and invoke the slash commands.
3. Configure the Tauri app – either ask the assistant to run /setup (it will propose Cargo.toml changes and plugin registration) or follow the manual guide.
4. Run the server – npx @hypothesi/tauri-mcp-server starts a stdio‑based MCP that talks to the assistant.
5. Interact – From the assistant type commands like /select (visual element picker) or /fix-webview-errors. You can also call the CLI directly, e.g. tauri-mcp webview-screenshot --file out.png.

Architecture at a glance

AI Assistant  ←→  MCP Server (Node.js)  ←WebSocket→  Tauri app
   │                │                     │
   │                ├─ Manager (CLI/config)   └─ MCP Bridge plugin (Rust)
   │                ├─ Driver (UI automation)   └─ Webview (DOM/UI)
   │                └─ Monitor (logs & IPC)    

The bridge plugin injects a small Rust library into the Tauri binary, exposing IPC commands and state that the server forwards to the assistant. No external browsers or drivers are required.

Getting started quickly

# Clone & install
git clone https://github.com/hypothesi/mcp-server-tauri.git
cd mcp-server-tauri
npm install

# Build packages
npm run build

# Start the server (will listen on a stdio pipe for the AI client)
npx @hypothesi/tauri-mcp-server

For a hands‑on demo you can also run the bundled test-app Tauri project.

Who should care

• Developers building desktop apps with Tauri who want AI‑driven debugging or code generation.
• Teams that use Claude, Cursor, or similar assistants and need the assistant to see the UI state.
• Anyone looking to script UI interactions across desktop and mobile Tauri targets without Selenium/Playwright.

License – MIT (see LICENSE).

---

The project is maintained by the hypothesi organization and is not officially affiliated with the Tauri Programme.