mcpField
M
MCP Sentry Server (Node.js)

MCP Sentry Server (Node.js)

Local
github.com/weekitmoDeveloper Tools1Updated 82w ago

About MCP Sentry Server (Node.js)

MCP Sentry Server (Node.js) connects MCP clients to Sentry issue data, including reports and stack traces. It offers both stdin/stdout and SSE transports, with tools and prompts for retrieving issues.

Key features

  1. 01Retrieves Sentry issues and stack traces
  2. 02Supports stdin/stdout and SSE
  3. 03Provides issue retrieval tools and prompts
  4. 04Configurable SSE port

Use cases

  1. 01Inspect a Sentry issue by ID or URL
  2. 02Find the most-triggered issue from a list
  3. 03Connect through an SSE endpoint

MCP support

The server exposes Sentry issue retrieval tools and prompts over MCP, using either stdin/stdout or SSE transport.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

No install method listed yet

We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.

Repository

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

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

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:

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
pnpm build
Run the Server
pnpm start

Alternatively, you can specify the Sentry authentication token and API base as command-line arguments:

pnpm start --auth-token your_sentry_auth_token_here --api-base your_sentry_api_base_here
Run with SSE Support
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
pnpm dev
MCP Configuration
{
  "mcpServers": {
    "sentry-server": {
      "command": "npx",
      "args": ["-y","@weekit/[email protected]"],
      "env": {
        "SENTRY_TOKEN": "your_sentry_auth_token_here",
        "SENTRY_API_BASE": "https://your_sentry_api_base_here"
      }
    }
  }
}

Using local build:

{
  "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:

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:

pnpm test

Run unit tests:

pnpm test test/unit.test.ts

Run integration tests:

pnpm test test/integration.test.ts

View test coverage:

pnpm test -- --coverage

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.