Sentry MCP Server
A Model Context Protocol server that provides AI assistants with comprehensive, read-only access to the Sentry API. Works with both Sentry SaaS and self-hosted instances.
What can it do?
Ask your AI assistant questions like:
- "What are the most frequent unresolved issues in the backend project?"
- "Show me the latest event for issue
PROJ-42with its full stack trace." - "List all projects in our Sentry organization."
- "What breadcrumbs led up to the latest crash in the mobile app?"
- "Give me the details of the event that triggered this alert."
- "How many unresolved issues do we have in the
api-gatewayproject?"
Features
- 8 read-only tools across 4 categories (Organizations, Projects, Issues, Events)
- Self-hosted support — point to any Sentry instance via
SENTRY_BASE_URL - LLM-optimized responses — internal fields are stripped, stack frames and breadcrumbs are truncated to save tokens
- No SDK dependency — lightweight
fetch-based HTTP client - Type-safe — built with TypeScript and Zod schema validation
- Security-first — read-only by design, supports scoped auth tokens
Supported Resources
| Category | Tools |
|---|---|
| Organizations | list_organizations |
| Projects | list_projects |
| Issues | list_issues, get_issue, get_issue_events |
| Events | list_events, get_event, get_latest_event |
Prerequisites
- Node.js >= 20
- A Sentry auth token with read scopes. You can create one at Sentry Settings > Auth Tokens (or the equivalent page on your self-hosted instance).
Key Differentiator: Self-Hosted Support
Unlike most Sentry integrations, this server fully supports self-hosted Sentry instances. Set the SENTRY_BASE_URL environment variable to your instance URL and the server will route all API calls there instead of sentry.io.
SENTRY_BASE_URL=https://sentry.yourcompany.com