mcp-sentry

# MCP Sentry Server (Node.js) This is a Node.js + TypeScript implementation of the Model Context Protocol (MCP) Sentry server. It provides integration with Sentry for retrieving error reports and stacktraces through MCP. [中文文档](./README_CN.md) ## Features - Integration with Sentry API for error tracking and monitoring - Support for both standard MCP over stdin/stdout and SSE (Server-Sent Events) transport - Configurable port for the SSE server (default: 3579) ## Prerequisites - Node.js >= 20.0.0 - pnpm >= 10.5.2 ## Installation Clone the repository and install dependencies: ```bash pnpm install ``` ## Configuration Create a `.env` file in the root directory with your Sentry authentication token: ``` SENTRY_TOKEN=your_sentry_auth_token_here ``` You can obtain your Sentry authentication token from your Sentry account settings. ## Usage ### Build the Project ```bash pnpm build ``` ### Run the Server ```bash pnpm start ``` Alternatively, you can specify the Sentry authentication token and API base as command-line arguments: ```bash pnpm start --auth-token your_sentry_auth_token_here --api-base your_sentry_api_base_here ``` ### Run with SSE Support ```bash pnpm start:sse ``` This will start an Express server with SSE support on port 3579 (default). You can access the SSE endpoint at: ``` http://localhost:3579/sse ``` You can also customize the environment variables using a `.env.local` file. ### Development Mode ```bash pnpm dev ``` ### MCP Configuration ```json { "mcpServers": { "sentry-server": { "command": "npx", "args": ["-y","@weekit/mcp-sentry@1.0.9"], "env": { "SENTRY_TOKEN": "your_sentry_auth_token_here", "SENTRY_API_BASE": "https://your_sentry_api_base_here" } } } } ``` Using local build: ```json { "mcpServers": { "my-sentry": { "command": "node", "args": ["path/to/dist/index.js"], "env": { "SENTRY_TOKEN": "your_sentry_auth_token_here", "SENTRY_API_BASE": "https://your_sentry_api_base_here" } } } } ``` ## Communication Protocols The MCP Sentry server supports two communication protocols: 1. **Standard MCP Protocol**: Communication over standard input/output streams 2. **SSE Transport**: Server-Sent Events for web-based communication via HTTP (port 3579) ## MCP Features The MCP Sentry server provides the following features through the MCP protocol: ### Prompts - `sentry-issue`: Retrieve a Sentry issue by ID or URL - `most-triggered-issue`: Find the issue affecting the most users from a Sentry issues list URL ### Tools - `get_sentry_issue`: Retrieve and analyze a Sentry issue by ID or URL - `get_sentry_issues_list`: Retrieve and analyze a list of Sentry issues from a URL ## API The server communicates using the Model Context Protocol over standard input/output streams and provides the following functionalities: 1. Listing available prompts and tools 2. Retrieving Sentry issue information including: - Issue title and ID - Status and severity level - First and last seen timestamps - Event count - Detailed stacktrace ## MCP Workflow Below is the workflow diagram of the MCP Sentry server: ```mermaid flowchart TD A[Client/LLM] -->|1. Send MCP Request| B[MCP Sentry Server] B -->|2. Parse Request Type| C{Determine Request Type} C -->|Prompt Request| D[Handle Prompt\nsentry-issue] C -->|Tool Request| E[Handle Tool\nget_sentry_issue] D -->|3. Extract Issue ID| F[Call Sentry API] E -->|3. Extract Issue ID| F F -->|4. Get Issue Data| G[Sentry API] G -->|5. Return Issue Data| F F -->|6. Get Event Data| G G -->|7. Return Event Data| F F -->|8. Parse Data| H[Create SentryIssueData Object] H -->|9. Format Data| I{Response Type} I -->|Prompt Response| J[Convert to PromptResult Format] I -->|Tool Response| K[Convert to ToolResult Format] J -->|10. Return Response| B K -->|10. Return Response| B B -->|11. Send MCP Response| A ``` ## Testing Run all tests: ```bash pnpm test ``` Run unit tests: ```bash pnpm test test/unit.test.ts ``` Run integration tests: ```bash pnpm test test/integration.test.ts ``` View test coverage: ```bash pnpm test -- --coverage ```