sentry-mcp

What is sentry-mcp?

Sentry's MCP service is a remote MCP server that acts as middleware to the upstream Sentry API, designed for human-in-the-loop coding agents. It is optimized for developer workflows and debugging use cases, and works with coding assistants like Cursor, Claude Code, and similar development tools.

How to use sentry-mcp?

The simplest way is to use the deployed service at https://mcp.sentry.dev. For Claude Code, install as a plugin: claude plugin marketplace add getsentry/sentry-mcp then claude plugin install sentry-mcp@sentry-mcp. For stdio transport, run npx @sentry/mcp-server@latest --access-token=sentry-user-token with optional --host for self-hosted instances and --insecure-http for plain HTTP. Remote clients can connect via a JSON configuration with a Sentry-Bearer header or through OAuth.

Key features of sentry-mcp

• Acts as middleware to the upstream Sentry API.
• Designed for developer debugging workflows and coding agents.
• Supports Claude Code plugin with automatic subagent delegation.
• AI-powered search tools for natural language queries.
• Stdio transport for self-hosted Sentry instances.
• OAuth and direct token authentication mechanisms.

Use cases of sentry-mcp

• Debugging Sentry errors, issues, traces, and performance from within a coding assistant.
• Using natural language to search Sentry events and issues via AI-powered tools.
• Integrating Sentry data into Claude Code, Cursor, or similar development tools.
• Monitoring and triaging errors on self-hosted Sentry deployments.

FAQ from sentry-mcp

What is sentry-mcp and how is it different from the general Sentry API?

Sentry-mcp is a remote MCP server that acts as middleware to the Sentry API, optimized for human-in-the-loop coding agents. Its tool selection and priorities are focused on developer workflows and debugging, not providing general-purpose Sentry functionality.

What are the runtime requirements for AI-powered search tools?

AI-powered search tools (search_events, search_issues, etc.) require an LLM provider (OpenAI, Azure OpenAI, Anthropic, or OpenRouter) and the EMBEDDED_AGENT_PROVIDER environment variable set explicitly. Without a configured provider, these specific tools are unavailable, but all other tools function normally.

How does data flow and authentication work for the remote transport?

The remote transport uses a Cloudflare worker that forwards the upstream Sentry API token via a custom Sentry-Bearer header. The worker does not store, validate, exchange, or refresh the token; the client or upstream provider remains responsible for token lifetime and refresh. OAuth is also supported.

Can I run sentry-mcp with self-hosted Sentry?

Yes. Use the stdio transport with the --host flag pointing to your self-hosted hostname. Add --insecure-http if the deployment only exposes plain HTTP. Some features (like Seer) may not be available; you can disable specific skills with --disable-skills=seer.

What transports and authentication methods are supported?

Sentry-mcp supports a remote transport (via Cloudflare) with OAuth or a direct Sentry token (via a Sentry-Bearer header), and a stdio transport using a Sentry user auth token. Additionally, it can be installed as a Claude Code plugin.