Sentry MCP
Sentry's official MCP server: search issues, inspect events and traces and analyze root causes from a coding agent
Install
claude plugin marketplace add getsentry/sentry-mcp
claude plugin install sentry-mcp@sentry-mcpAdds the sentry-mcp subagent.
This is third-party code. Review the repository files before installing.
What it does
The server is a layer over the Sentry API tuned for debugging in coding agents rather than full administration. The agent can search and inspect issues, view events and performance traces, and run root cause analysis through Seer. Natural-language search over events and issues is translated into Sentry query syntax by an embedded LLM agent. The main option is the remote server at mcp.sentry.dev with OAuth sign-in, and a stdio mode via npx serves self-hosted Sentry. In Claude Code there is also a plugin with a subagent that Claude delegates error questions to.
Who it is for. For developers whose application errors and performance data live in Sentry.
Good fit when
- You need to find and inspect a fresh production error without leaving the editor
- You want to connect a slow request trace to the project code
- You run self-hosted Sentry and want to connect an agent to it
Not a fit when
- You need full administration of a Sentry organization
- Your errors are collected by another monitoring system
Example request
Find the most frequent errors in the api project in Sentry over the last day and propose a fix for the top oneLimitations
You need a Sentry account, and the stdio mode needs a token with write scopes on projects and teams. Natural-language search tools need an OpenAI, Azure OpenAI, Anthropic or OpenRouter key; without one only the other tools work. Seer may be unavailable on self-hosted installs. The authors describe the stdio mode as a work in progress. OpenAI and Anthropic APIs are not reachable from Russia without a VPN.
How to disable. Remove the sentry server from the MCP config or run claude plugin uninstall sentry-mcp@sentry-mcp. Revoke the Sentry token in account settings.
MCP
- Transport
- http
- Authentication
- API key
| Environment variables | |
|---|---|
| SENTRY_ACCESS_TOKEN required, secret | Sentry user auth token for the stdio mode. |
| EMBEDDED_AGENT_PROVIDER | LLM provider for natural-language search: openai, azure-openai, anthropic or openrouter. |
| OPENAI_API_KEY secret | OpenAI key when that provider is used. |
| ANTHROPIC_API_KEY secret | Anthropic key when that provider is used. |
| OPENROUTER_API_KEY secret | OpenRouter key when that provider is used. |
| SENTRY_HOST | Self-hosted Sentry host; cloud Sentry is used when unset. |
| MCP_DISABLE_SKILLS | Comma-separated tool groups to disable, for example seer. |
Security check
- The token grants write access to Sentry projects, teams and events
- Error data may be sent to a third-party LLM provider for search
README in short
The README explains that the main path is the hosted service at mcp.sentry.dev and shows the Claude Code plugin install. It covers the stdio mode for self-hosted Sentry: token scopes, --host and --insecure-http flags, environment variables and an MCP config example. It also describes passing a token to the remote server via a header. The rest covers local development, tests and evals.
FAQ
Can I disable some tools?
Yes, with the --disable-skills flag or the MCP_DISABLE_SKILLS variable, for example to drop Seer.
Can I pass a Sentry token to the remote server without OAuth?
Yes, clients that support headers can send it in the Authorization header with the Sentry-Bearer scheme.
Related
An MCP server built into the Netdata agent: metrics, logs, alerts and live process, service and container data for an AI assistant
GitHub's official MCP server: code, issues, pull requests, Actions and security alerts straight from the agent
AWS's official MCP server suite: docs, IaC, containers, serverless, databases, cost and monitoring
A GitHub Action that runs Claude Code in pull requests and issues: answers mentions, reviews and makes changes