MCP Server Tauri
An MCP server that gives an agent screenshots, DOM, logs and IPC of a running Tauri v2 app and lets it click through the interface
- 329 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The agent runs JavaScript in your app's webview, types text and calls IPC commands
- Screenshots, DOM and logs of the app land in the model context
Install
- Works in
Manual install
npx -y @a1st/aix add mcp tauri --command 'npx @hypothesi/tauri-mcp-server' --useraix detects installed agents and writes the server into their configuration. Restart the agent after the command.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent runs JavaScript in your app's webview, types text and calls IPC commands
- Screenshots, DOM and logs of the app land in the model context
This is third-party code. Review the repository files before installing.
What it does
MCP Server Tauri connects an agent to a running Tauri v2 app. The Node.js server talks to the agent over stdio and to the app over a WebSocket on port 9223. Inside the app runs the tauri-plugin-mcp-bridge plugin written in Rust, which you add to the project through Cargo.toml. The server exposes 21 tools. The agent takes a webview screenshot, finds elements by selector, clicks and types, and can also run JavaScript in the webview. There is an accessibility tree snapshot and a visual element picker: you click in the app and the agent gets metadata and a screenshot. For the backend there are IPC command calls, IPC traffic capture and event emission. Logs come from the webview console, the system and mobile platforms. Three ready commands come with it. `/setup` wires the plugin into the project, `/fix-webview-errors` finds and fixes JavaScript errors in the webview, and `/select` returns the element you picked. A window is chosen with the windowId parameter.
Who it is for. For developers of desktop and mobile apps on Tauri v2 who want the agent to drive the interface, look at it and verify its own changes.
Good fit when
- The agent edits a Tauri app's frontend and needs to see the result in a screenshot
- JavaScript errors appear in the webview and you need a `/fix-webview-errors` run that reads the console
- You want to check which IPC commands the interface calls: the server records IPC traffic and hands it to the agent
- You want to point the agent at a specific interface element with a mouse click
- The app builds for iOS and Android and you need simulator or emulator logs
Not a fit when
- The project is on Tauri 1 or on Electron: the plugin and server are built for Tauri v2
- You test a regular website in a browser: Playwright and other browser tools cover that
- The app holds data you must not show an agent: screenshots, DOM and logs land in the model context
- The machine has no Node.js 20, Rust and Cargo, and you cannot install the Tauri CLI
Example request
Run the app, take a screenshot of the main window and find out why the save button does not respond. Read the webview console.MCP
- Transport
- stdio
- Authentication
- not required
Limitations
- Needs Node.js 20 or newer, Rust with Cargo and the Tauri CLI (`npm install -g @tauri-apps/cli@next`). Mobile work also needs Xcode on macOS or the Android SDK.
- The app must run with the tauri-plugin-mcp-bridge plugin connected: without it the server gets no DOM or IPC.
- Projects on Tauri 1 are not supported.
- The author warns that the server was written with agentic coding tools and may contain bugs.
- The command-line client for work without an agent is a separate package, @hypothesi/tauri-mcp-cli.
How to disable. Remove the tauri entry from your agent's MCP configuration and restart it. Remove the tauri-plugin-mcp-bridge plugin from Cargo.toml and from the app registration if you no longer need it.
FAQ
How do I install MCP Server Tauri in Claude Code?
Run `npx -y @a1st/aix add mcp tauri --command 'npx @hypothesi/tauri-mcp-server' --user`. The aix utility detects installed agents on its own, including Claude Code, Cursor and Codex. Then restart the agent.
What do I add to a Tauri app?
The tauri-plugin-mcp-bridge plugin. It is easier to ask the agent with the phrase "Help me set up the Tauri MCP Bridge plugin" or call `/setup`: it examines the project, shows the Cargo.toml changes and asks permission. Manual setup has its own guide.
Can I drive the server from a terminal without an agent?
Yes, there is a command-line client for that: `npm install -g @hypothesi/tauri-mcp-cli`. Then commands such as `tauri-mcp driver-session start --port 9223` and `tauri-mcp webview-screenshot --file screenshot.png` work.
Does the server work with several windows?
Yes. All webview tools accept an optional windowId parameter, and the manage_window tool with the list action shows available windows.
What can the server do on mobile platforms?
The list_devices tool lists iOS simulators and Android emulators. The read_logs tool reads Android logcat and iOS logs, and gestures include swipe and long press.
Discussion
No questions or reviews yet.
Related
MCP serversA CLI and TypeScript runtime for calling MCP servers from a terminal, scripts and generated typed clients
A plugin with skills for building TypeScript MCP servers and MCP apps: the agent scaffolds, edits, debugs and verifies them
An MCP server that lets an agent read files, search code and run commands in a workspace
An MCP server and CLI for Siemens TIA Portal V20 and V21: the agent builds a project with PLC and HMI screens and compiles it
A local code graph with version history: an MCP server that tells the agent who calls a function and what an edit breaks
An MCP server that builds a code graph once and serves the answer to the agent instead of re-reading files
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| MCP Server Tauri (this entry) | no | not required | 329 | |
| Argent | no | 3,127 | drives iOS simulators, Android emulators, TVs and Electron apps | |
| Playwright CLI | yes | 14,754 | drives a browser with short commands and fits websites |
README in short
The tools are grouped into tables, and a three-layer diagram shows the path from the agent through the server to the app. The repository is a monorepo with the server, the plugin and a test app.