What does @sentry/mcp-server actually do?

It lets your AI assistant (Claude Code, Cursor, etc.) query your Sentry project directly — reading real errors, stack traces, and performance data without you copy-pasting anything.

Why do most setups fail?

Three common reasons: wrong token type (User Auth Token instead of Internal Integration), missing scopes, or the editor not picking up env vars correctly. This guide covers all three.

What's included in the Notion template?

8 sections: What is @sentry/mcp-server, Auth token setup, Claude Code config, Cursor/Windsurf/Cline config, production env vars, 30+ power prompts, 20+ troubleshooting entries, and advanced AI agent pipeline integration.

44,000 downloads/week — @sentry/mcp-server

Your Sentry MCP config
is probably broken.

Name: Sentry MCP Setup — Production Guide
Brand: Solia
Price: 9.00 EUR
Availability: InStock

Most developers install @sentry/mcp-server with the wrong token type, missing scopes, or in an editor that ignores env vars. This guide fixes all three — and gives you 30+ prompts to actually use it.

Get the Notion template — 9€ One-time payment · Instant access · 8 sections · 30+ prompts

If you're building with AI agents, you know this:

You copy-paste stack traces from Sentry into Claude Code. Every. Single. Time.
You tried setting up the MCP but the AI still says "I don't have access to Sentry".
You spent 40 minutes debugging a 403 error — wrong token type, obviously.
Your agent pipeline crashes in production and you have no idea which step failed.
The official Sentry docs show the package exists but nothing about production config.

What's inside the Notion template

What is @sentry/mcp-server

How it works, what data it can access, and why most devs use it wrong.

Create the right auth token

Internal Integration vs User Token. Exact scopes needed. Step-by-step with screenshots.

Setup in Claude Code

Project-level + global config. Verification steps. How to confirm the MCP is live.

Cursor / Windsurf / Cline / VS Code

Config file locations for every major editor. Copy-paste ready JSON.

Production config

Version pinning, multi-org setup, secrets management, self-hosted Sentry.

30+ power prompts

Bug investigation, release analysis, performance, agent debugging, batch triage.

Troubleshooting — 20+ errors

Every 401, 403, "not found", "crashes immediately" — with exact fixes.

AI agent pipeline integration

How to instrument agents with Sentry tags, close the code→error→fix loop.

How it works

Buy the template (9€, one-time)

Instant access via Gumroad. You get a Notion template link — duplicate it into your workspace in 30 seconds.

Follow the setup guide for your editor

Claude Code, Cursor, Windsurf, Cline — each has a dedicated section with copy-paste config. Takes 5-10 minutes.

Use the prompts — your AI reads your real errors

Instead of copy-pasting stack traces, you just ask: "Look at the last 5 errors in project X and propose fixes."

Sample prompts included

30+ prompts, ready to paste into Claude Code or Cursor.

List the top 10 unresolved errors in [project-slug] sorted by frequency.
For the #1 error, show the full stack trace and last 5 breadcrumbs.

Find all errors in [project-slug] that occurred in the last 24h
and affect more than 5 users. Group them by root cause.

My AI agent pipeline is throwing errors tagged environment="production".
Build a retry strategy based on the error patterns.

+ 27 more prompts: release analysis, performance bottlenecks, agent debugging, batch triage, CLAUDE.md integration snippet. All included.

What this is not

Not a Sentry tutorial for beginners — assumes you already have a project with errors.
Not a subscription — 9€ once, no renewal, no account required.
Not specific to one editor — covers Claude Code, Cursor, Windsurf, Cline, and VS Code.
Not a wrapper or custom MCP server — uses the official @sentry/mcp-server package.

Who made this

👨‍💻

Damien Reiss — Solia

I build small tools for developers using AI coding assistants. I spent 2 hours debugging an @sentry/mcp-server setup that failed because I used a User Auth Token instead of an Internal Integration. The official docs don't mention this clearly. This guide does.

Solia makes small tools for developers — nothing complex, nothing behind "book a demo". Just things that save you time.

Questions

Why does my Sentry MCP say "authentication failed" even with a valid token?

Most likely you created a User Auth Token instead of an Internal Integration token. These look similar but have different scopes. The guide includes a step-by-step to create the right type, and how to verify your token type before debugging further.

Does this work with self-hosted Sentry?

Yes. The production config section covers the SENTRY_HOST environment variable, SSL certificate handling for internal CAs, and common issues with self-hosted instances behind VPNs or reverse proxies.

I set it up but the AI still doesn't fetch Sentry data. What's wrong?

Three most likely causes: (1) the editor didn't pick up env vars — restart completely, not just reload window. (2) The MCP process is running but returning empty results — check your project slug matches exactly. (3) You're in a new conversation — the AI starts fresh each time. The troubleshooting section covers 20+ error patterns with exact fixes.

Can I share this with my team?

It's a personal license (one person). For teams of 3+, buy one copy per person. If you want a team license (5+ people), email contact@solia.me — I'll set it up.

What if it doesn't help me?

Email me within 14 days for a full refund. No questions, no process. contact@solia.me.

Your Sentry MCP config is probably broken.