Sentry MCP Server: Setup Guide, Use Cases & Limitations 2026

What Is the Sentry MCP Server?

The Sentry MCP Server connects AI coding assistants directly to your Sentry error monitoring account via the Model Context Protocol. Instead of switching tabs to investigate a crash, you query Sentry issues, stack traces, and project data from inside Claude, Cursor, or any MCP-compatible client. This sentry mcp server setup guide covers everything from install to known gaps.

Sentry monitors over 4 million projects and processes billions of error events per month — making its data one of the highest-signal debugging inputs an AI assistant can access. Pairing that signal with an LLM inside your editor (see our Cursor Review 2026 for context on how MCP tools slot in) cuts context-switching time to near zero.

Quick Facts

What You Can Do With It

• Fetch issues by ID or project — pull full issue details, event counts, and first/last seen timestamps directly into your chat context
• Read stack traces — retrieve raw stack frames, exception types, and breadcrumbs without leaving your editor
• List projects and organizations — enumerate all Sentry projects your token has access to
• Query issues by filter — filter by status (unresolved, resolved, ignored), assignee, or tag
• Resolve and assign issues — update issue status or assignee from the AI client
• Search events — run Sentry’s full query syntax (is:unresolved assigned:me) through the MCP interface

This pairs naturally with other debugging-adjacent MCP servers — the Jira MCP Server handles ticket creation while Sentry MCP handles the raw error data.

How to Set It Up

Step 1 — Generate a Sentry Auth Token

Go to https://sentry.io/settings/account/api/auth-tokens/ → Create New Token → grant scopes: project:read, event:read, issue:read, issue:write (write only if you want resolve/assign actions).

Step 2 — Install and run

npx @sentry/mcp-server

No global install required. npx pulls the latest version on each run.

Step 3 — Configure your MCP client

For Claude Desktop, add to claude_desktop_config.json:

{
  "mcpServers": {
    "sentry": {
      "command": "npx",
      "args": ["@sentry/mcp-server"],
      "env": {
        "SENTRY_AUTH_TOKEN": "your_token_here",
        "SENTRY_ORG": "your-org-slug"
      }
    }
  }
}

For Cursor or Windsurf (see our Windsurf Review 2026 for MCP config location), the same JSON block applies under their respective MCP settings files.

Step 4 — Verify connection

Ask your AI client: “List my Sentry projects.” A successful response returns project slugs and IDs within 2–3 seconds.

Works Best With

The Supabase MCP Server and Sentry MCP together cover database errors plus application errors — a common pairing for full-stack debugging sessions.

Known Limitations

• No performance data — Sentry Performance (transactions, spans, Web Vitals) is not exposed through the MCP interface as of 2026
• No replays — Session Replay data is unavailable via MCP tools
• No alerts management — you cannot create or modify alert rules through the server
• Rate limits apply — Sentry API rate limits (varies by plan) apply to all MCP calls; high-frequency queries will hit throttling
• Self-hosted Sentry requires extra config — you must set SENTRY_URL to your instance URL; not documented in the default README
• Token scope errors are silent — insufficient token scopes return generic errors with no scope-specific message

Community Resources

• GitHub repo: getsentry/sentry-mcp — official issues tracker and changelog
• Sentry Discord: #mcp-integrations channel in the official Sentry Discord server
• MCP specification: modelcontextprotocol.io — reference for tool schema and transport layer
• Sentry API docs: https://docs.sentry.io/api/ — full endpoint reference for understanding what data the MCP server exposes
• Related MCP servers: Linear MCP Server for issue tracking, Slack MCP Server for alert routing

FAQ

Does the Sentry MCP Server work with self-hosted Sentry installations?
Yes. Set the SENTRY_URL environment variable to your self-hosted instance URL (e.g., https://sentry.yourcompany.com). The default points to sentry.io.

Which token scopes are required for read-only access?
Minimum required: project:read and event:read. Add issue:write only if you need resolve or assign actions from the AI client.

Does it support Sentry’s full query syntax?
Yes. You pass standard Sentry query strings (e.g., is:unresolved assigned:me) and the server forwards them to the API unchanged.

Can I use this with multiple Sentry organizations?
Not in a single server instance. Run separate server instances with different SENTRY_ORG values and register each as a distinct MCP server in your client config.

Is the Sentry MCP Server free to use?
The server itself is open-source and free. Sentry API access is governed by your Sentry plan — the free Developer plan includes limited API rate limits; Team and Business plans increase those limits.