Official Dev Tools

Sentry

Pull errors, issues, and performance data from Sentry.

Works with: Claude DesktopClaude CodeCursorWindsurf
Quick install
npx -y @sentry/mcp-server

How to install the Sentry MCP server

Add this to your Claude Desktop MCP configuration:

{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": [
        "-y",
        "@sentry/mcp-server"
      ]
    }
  }
}

Add this to your Claude Code MCP configuration:

npx -y @sentry/mcp-server

Add this to your Cursor MCP configuration:

{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": [
        "-y",
        "@sentry/mcp-server"
      ]
    }
  }
}

Add this to your Windsurf MCP configuration:

{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": [
        "-y",
        "@sentry/mcp-server"
      ]
    }
  }
}

The Sentry MCP server gives Claude full read access to your error tracking and performance data. For any team running Sentry in production, this is one of the highest-leverage installs available.

The combination of Sentry plus Claude Code is genuinely powerful. An error fires in production, Claude reads the stack trace, pulls the relevant code from the GitHub MCP server, identifies the introducing commit, and drafts a fix. End-to-end debugging from a single prompt.

Why use it

Most production debugging follows the same loop: see the error in Sentry, find the relevant code on GitHub, write a fix, deploy. Each step is a tab switch and a copy-paste. The MCP server collapses the whole loop.

For solo founders running production systems, this is the closest thing to having a senior engineer on call for incident response. Claude won’t catch every bug, but it’ll handle the obvious ones without you needing to context-switch.

What it actually does

A wide surface: list issues with filters (project, environment, level, status), fetch a specific issue with its events and breadcrumbs, search across organisations, pull performance metrics, list releases and their associated issues. With write scopes it can resolve issues, ignore them, and add comments.

Practical patterns:

  • “What’s the highest-frequency error in production right now?”
  • “Show me every issue first seen after the v0.1.6 release.”
  • “Pull the stack trace for issue PROJ-123 and find the introducing commit.”

Gotchas

Token scoping matters. A token with org:admin can do a lot of damage if Claude misinterprets a prompt. Default to read-only and add issue:write only when you specifically need triage actions.

Issue volume can be huge. A query like “all unresolved issues” against a noisy project may return thousands. Use filters aggressively (level, environment, time range) or the response will be too large for Claude to reason about cleanly.

Pair with GitHub for the full debug flow: Sentry surfaces the error, GitHub finds the introducing PR, Claude drafts a fix. If you also have PostgreSQL installed, you can verify the fix against production data in the same conversation.

MCP servers related to Sentry

All in Dev Tools →

GitHub

Anthropic's official GitHub server: repos, issues, PRs, code search.
F

Filesystem

Anthropic's official local filesystem server with read, write, and search.
G

Git

Anthropic's official Git server for local repository operations.

GitLab

Anthropic's official GitLab server for repos, MRs, and pipelines.

Sentry MCP server: FAQs

Is the Sentry server official?

Yes. Sentry ships and maintains the @sentry/mcp-server package. It's actively developed and gets feature parity with the dashboard relatively quickly.

Does it work for Sentry self-hosted?

Yes. The server takes a base URL config that points at your Sentry instance, whether that's sentry.io or a self-hosted deployment. Same API, same features.

What permissions does it need?

An auth token from your Sentry organization with at least the project:read and event:read scopes. Add issue:write if you want Claude to resolve or comment on issues.

Can it triage issues automatically?

Yes, that's one of the main reasons to install it. Common pattern: feed Claude an issue ID, it pulls the stack trace and breadcrumbs, then searches GitHub via the GitHub MCP server to find the introducing PR. Combined with Claude Code, you can go from production error to fix-PR in a single conversation.

Does it support Sentry's AI-suggested fixes?

Sentry's own AI features remain in the Sentry UI. The MCP server exposes the underlying issue data, which Claude can reason about independently. If Sentry's AI suggestion is useful, Claude can also pull it via the API.

Back to all MCP servers