MCP Server Tauri connects an AI assistant to a running Tauri v2 application so it can build, test, and debug it. Screenshots, accessibility-tree snapshots, console logs, and captured inter-process communication give the assistant a view of the app's actual state, and a matching set of tools lets it click, type, scroll, wait for elements, execute JavaScript, and call Tauri commands directly. It is an unofficial community project, not affiliated with or endorsed by the Tauri Programme.
Two pieces make it work. A Node.js MCP server speaks the protocol over stdio to the assistant and holds the manager, driver, and monitor components. A Rust bridge plugin, registered in the Tauri application under debug builds, connects back over a WebSocket and exposes the app's IPC commands, events, and backend state. The app must expose the global Tauri object for the bridge to reach it. Optional TypeScript bindings exist for calling the plugin from frontend code, but the server talks to the Rust plugin directly. Because the bridge is native, no external browser driver such as Selenium or Playwright is involved, and the same setup works on Linux, Windows, macOS, Android, and iOS.
The same tools are reachable from a command-line client for terminal, script, and continuous-integration use, with keep-alive support so an automation session survives across separate commands, and agent skills so coding agents learn the correct usage. Three guided slash commands cover setting up or updating the bridge plugin, finding and fixing JavaScript errors in the webview, and visually selecting an element to return its metadata and a screenshot.
Features
- UI automation: screenshots, clicks, scrolls, swipes, long-presses, typing and key events, element finding by selector, computed style reads, structured accessibility-tree snapshots, and waiting for elements, text, or events
- Visual element picking: a picker that returns metadata and a screenshot for an element the user clicks, plus metadata for an element selected with a keyboard modifier
- IPC monitoring: start and stop monitoring, read captured Tauri IPC traffic, execute IPC commands, emit custom events, and read app metadata and backend state
- Logs: stream console output, Android logcat, iOS, and system logs
- Multi-window support: every webview tool takes an optional window identifier, and a window tool lists, inspects, and resizes windows
- Mobile development: list connected Android devices and iOS simulators
- Two entry points: an MCP server for MCP-capable clients and a command-line client for terminals, scripts, and continuous integration
- Slash commands: guided workflows for bridge-plugin setup, webview error fixing, and visual element selection
