What for: give AI the context of real errors and traces from Sentry so you can debug production issues right in your IDE.

What it gives the AI

The AI gets the context of real production errors from Sentry: the issue list, stack traces, frequency, affected releases and performance traces. You can ask right in your IDE "what broke in prod and why" — and the AI proposes a fix based on the actual error rather than a guess. There's also the Seer AI agent for automated root-cause analysis.

Example: "Look into the most frequent error from the last 24 hours" — the AI takes the top issue by frequency, reads the stack trace and points at the line and the cause with a suggested patch.

Setup (exact steps)

The official hosted remote server with OAuth — no manual tokens needed.

Claude Code

claude mcp add --transport http sentry https://mcp.sentry.dev/mcp

Then /mcp → sign in to your Sentry account in the browser.

Cursor / Claude Desktop

{
  "mcpServers": {
    "sentry": {
      "type": "http",
      "url": "https://mcp.sentry.dev/mcp"
    }
  }
}

Usage example

Say: "Show me the stack trace for error ID abc123 and suggest a fix" → the AI pulls the issue details from Sentry, works through the trace and proposes a patch. Or "which release introduced these new errors?" — it correlates errors with deploys.

Security

OAuth, least privilege: access is granted by signing in to Sentry; the AI reads errors and traces — a diagnostic, essentially read-oriented scenario (human-in-the-loop).
Organization/project scope: limit the scope on the Sentry side so the AI only sees the projects it should.
Revoking access: any time via /mcp → "Clear authentication".

Gotchas

This is an external hosted service (mcp.sentry.dev): for a closed perimeter or self-hosted Sentry, check for a separate connection option.
Stack traces and issue lists are bulky and run into the MCP output limit — ask about one specific error, not "everything from the past month".
Error data can contain PII from request context — keep that in mind before running it through an AI.

Sentry MCP