This is an official Sentry-built Model Context Protocol (MCP) server that exposes Sentry's error monitoring, issue tracking, and performance data as tools consumable by AI coding assistants. It is deployed as a remote service at mcp.sentry.dev and also supports a local stdio transport via npx.

It solves the problem of AI coding agents lacking direct access to runtime error data, stack traces, and performance issues captured in Sentry. Without this, developers must manually copy-paste Sentry data into their chat context; this server automates that bridge.

Primary users are developers using AI-assisted coding tools (Cursor, Claude Code, VS Code Copilot, etc.) who want their agent to automatically pull Sentry issues, events, and traces when debugging. It is also used by Sentry's own internal automation and CI pipelines via evaluation tests.

The architecture is a pnpm monorepo using Turborepo. The core logic lives in @sentry/mcp-core (tools, API client, skills system). The Cloudflare Workers deployment (@sentry/mcp-cloudflare) handles remote OAuth-backed access and hosts a React UI. The @sentry/mcp-server package handles stdio transport. AI-powered search tools (search_events, search_issues) embed a sub-agent that translates natural language into Sentry query syntax using an external LLM (OpenAI, Anthropic, etc.).

Notable: The project has an unusually sophisticated AI-agent development infrastructure including agent skill definitions, LLM evaluation harnesses (vitest-evals), AST-grep lint rules, and plugin manifests for multiple AI IDEs. The license is FSL-1.1-ALv2 (Functional Source License), not a standard open-source license—commercial use restrictions apply for competing products. The repository version is tagged 0.37.0 in workspace packages but 0.0.1 at root, suggesting the root package.json is a workspace coordinator, not a published artifact.