MCP Server Tauri

# Suggested Prompts for mcp-server-tauri These example prompts are designed for use with an AI assistant connected to this MCP server. They map directly to the tools and workflows described in the [API reference](/api/) and `AGENTS.md`. You can copy them as-is or adapt them to your project. --- ## 1. Project Setup & Configuration **Prompt:** "Help me set up the Tauri MCP bridge plugin in my existing Tauri v2 app. Please: - Inspect my repo structure and Tauri config files. - Tell me what files I need to add or update to install `@hypothesi/tauri-plugin-mcp-bridge`. - Update my Tauri config to register the plugin. - Add any required Rust code to `src-tauri` and explain where it goes. - Finish by telling me how to run the app so the MCP server can connect." **Prompt:** "Review my Tauri configuration and summarize anything relevant for MCP usage. Please: - Find my main Tauri config and plugin registration. - List which windows exist and how they are configured. - Identify anything that might interfere with the MCP bridge (e.g., CSP, dev server URLs). - Suggest any changes needed so UI automation and IPC monitoring will work reliably." --- ## 2. Mobile Devices & Simulators These prompts use the **mobile development** tools (see [/api/mobile-development](/api/mobile-development)). **Prompt:** "List all available Android emulators and iOS simulators on my machine. Tell me: - Which ones are currently running. - For each device: name, platform, and OS version. - Which device you recommend for testing this app and why." **Prompt:** "Start an Android emulator suitable for my Tauri app and confirm when it is ready for UI testing. Then: - Explain how you will connect the Tauri app to this device. - Describe any extra steps I need to take (e.g., running `npm run tauri android dev`)." --- ## 3. UI Automation & Visual Testing These prompts use the **UI automation** and **webview interaction** tools (see [/api/ui-automation](/api/ui-automation) and [/api/webview-interaction](/api/webview-interaction)). **Prompt:** "Connect to my running Tauri app and verify that the login flow works. Please: - Open the main window. - Type a test email and password into the login form. - Click the login button. - Wait for navigation or any success indicator. - Report whether the login succeeded and capture a screenshot of the resulting page." **Prompt:** "Find the primary call-to-action button on the home screen and verify it is visible and clickable. Steps: - Wait for the main window and root view to load. - Locate the button by its text or data attributes. - Verify it is enabled, visible, and not overlapped. - Take a screenshot and annotate what you interacted with. - Report any console errors that occurred while testing this interaction." **Prompt:** "Inspect the styles applied to the header navigation element. Please: - Locate the header nav bar element by selector. - Show me its computed CSS (colors, font, layout-related properties). - Explain anything that might cause layout issues or overflow on smaller screens." --- ## 4. JavaScript Errors & Console Debugging These prompts lean on console log and JavaScript execution tools. Unhandled errors and unhandled promise rejections in the webview are automatically logged via `console.error` with the prefixes `[MCP_UNHANDLED_ERROR]` and `[MCP_UNHANDLED_REJECTION]`, which you can access through `tauri_read_logs` with `source: "console"`. **Prompt:** "My Tauri app is throwing errors when I click a button, but I cannot see them clearly. Please: - Attach to the running webview. - Clear the console. - Click the problematic button. - Capture and summarize all console logs and stack traces. - Suggest likely causes and the next code locations I should inspect." **Prompt:** "Show me all unhandled errors and unhandled promise rejections from the webview in the last 5 minutes. Please: - Call `tauri_read_logs` with `source: "console"`. - Filter logs for `[MCP_UNHANDLED_ERROR]` or `[MCP_UNHANDLED_REJECTION]`. - Summarize the distinct error messages and how often they occurred. - Suggest where in the codebase I should start investigating for each one." **Prompt:** "Run a small JavaScript snippet in the main webview to confirm the app's runtime state. I want to know: - What the current route is. - Whether a specific global variable (or store entry) is set. - Whether any uncaught promise rejections are present in the console logs." --- ## 5. IPC Monitoring & Backend Debugging These prompts use the **IPC plugin** tools (see [/api/ipc-plugin](/api/ipc-plugin)). **Prompt:** "Set up IPC monitoring for my Tauri app and show me everything that happens when I open the settings page. Please: - Start an IPC monitor session. - Trigger the navigation to the settings page (via UI or command). - Capture all IPC invocations and events during that period. - Summarize which commands are called and with what arguments. - Point out any failures, timeouts, or suspicious patterns." **Prompt:** "I want to debug why a specific IPC command (for example `get_user_profile`) is failing. Please: - Start IPC monitoring. - Trigger the feature in the UI that calls this command. - Use `tauri_ipc_get_captured` (or equivalent) to retrieve the captured traffic. - Show the request and response payloads. - Suggest changes to either frontend or backend code to make error handling more descriptive." --- ## 6. E2E Testing with the Test App These prompts focus on `packages/test-app` and end-to-end flows. **Prompt:** "Create or update an end-to-end test in `packages/test-app` that validates IPC monitoring. The test should: - Launch the `test-app`. - Start a Tauri driver session and an IPC monitoring session. - Trigger a known IPC command from the frontend (e.g., `invoke('ping')`). - Assert that the IPC plugin captured the call with the expected payload. - Clean up by stopping the monitor session and closing the app. - Integrate with the existing test runner used in this repo." **Prompt:** "Review the existing end-to-end tests in `packages/test-app`. Please: - List the main flows currently covered. - Identify any missing coverage for mobile, UI automation, or IPC. - Propose 2–3 additional test cases that would significantly improve confidence in the MCP tools." --- ## 7. Implementing a New MCP Tool These prompts align with the workflow described in `AGENTS.md` for adding tools. **Prompt:** "Add a new tool to the MCP server that inspects the current Tauri window list. Requirements: - Follow the existing patterns in `packages/mcp-server/src/tools-registry.ts`. - Define a Zod schema for the input and output. - Implement the handler in the appropriate driver or manager module. - Wire it into the Rust plugin in `packages/tauri-plugin-mcp-bridge` if native support is required. - Add an E2E test in `packages/mcp-server/tests/e2e/`. - Document the new tool in the API docs." **Prompt:** "Review the implementation of a new MCP tool I just added. Please: - Check the TypeScript types and schema definitions. - Verify that the tool is registered correctly in `tools-registry.ts`. - Ensure the Rust side (`tauri-plugin-mcp-bridge`) is consistent with the TypeScript contract. - Suggest improvements for error handling, logging, and test coverage." --- ## 8. Error Handling & Resilience **Prompt:** "Review the error handling architecture in the `mcp-server` package, especially the Driver and Manager classes. Goals: - Find where errors are swallowed or turned into generic messages. - Ensure all tool failures return structured, descriptive error payloads. - Add or refine errors for cases where the Tauri app is closed or the WebSocket connection is lost, returning messages like `Tauri app not connected`. - Propose a consistent error format for all tools and update call sites accordingly." **Prompt:** "Simulate common failure scenarios for the MCP server. Please: - Test what happens if the Tauri app is closed while a session is active. - Test what happens if the WebSocket cannot be established. - Capture the current error messages returned to the AI client. - Suggest specific code changes to make these failures clearer and more actionable." --- ## 9. Release & Version Management These prompts follow the release process documented in `AGENTS.md`. **Prompt:** "Walk me through preparing a new release for this monorepo. Please: - Inspect the git log since the last tag to identify changes. - Propose the correct next version (patch, minor, or major) with justification. - Update all three changelogs (`CHANGELOG.md`, `packages/mcp-server/CHANGELOG.md`, `packages/tauri-plugin-mcp-bridge/CHANGELOG.md`). - Update the version fields in both `package.json` files and in `Cargo.toml`. - Explain which lock files need to be updated and how. - Draft the final commit message and tag name." **Prompt:** "Validate that all version numbers in this repo are consistent for the current release. Steps: - Check the `version` fields in the relevant `package.json` files. - Check the plugin's `Cargo.toml` and both `Cargo.lock` files. - Confirm that the changelogs contain entries for every version between the last tag and the current one. - Report any mismatches and suggest concrete edits." --- ## 10. General Development Workflow These prompts combine multiple tools for everyday development. **Prompt:** "Help me debug why my Tauri app's startup is slow. Please: - Start a Tauri driver session and attach to the main window. - Monitor IPC calls during startup and summarize what happens. - Capture console logs, network-like requests, and heavy operations. - Suggest concrete code-level optimizations based on what you see." **Prompt:** "Act as a Tauri development assistant. Right now I want to: - Build and run my app. - Verify that all windows render without errors. - Run a basic regression pass across the main pages (home, settings, about). - Capture screenshots for each page. - Report any console errors, failed IPC calls, or visual anomalies you detect."