Connects to your Sentry org and lets the model query issues, list events, fetch stack traces, and update issue status. Pairs naturally with the GitHub MCP for "find the regression, write the fix, open a PR" flows.

Pull error events and issue context into Claude for triage.

Use Case

"What's the most-impacting error this week" with full stack trace,Triaging unowned issues by inferring component from stack,Generating fix PRs from a Sentry event with full context,Periodic error report drafted into a Slack message

When Not to Use

Bugsnag, Rollbar, or other APMs — different servers,Real-time alerting (use Sentry's native alert rules, not the MCP),Read-only org access — write ops need user-level token

Common Pitfalls

  • Auth token must have project:read minimum; defaults are too narrow
  • Self-hosted Sentry needs SENTRY_HOST env var pointing at your URL
  • Issue list paginates at 25; ask for specific issue IDs for deep dives