sentry

This MCP server acts as a middleware layer between clients (like Claude Code or other coding assistants) and the upstream Sentry API, optimized for developer workflows and debugging tasks. It exposes a set of AI-assisted tools that can search and query Sentry data (errors, issues, traces, performance) and delegates appropriate tasks to the underlying MCP server bundled by Sentry. The provided example shows how to run the server via npx and how to configure authentication tokens and host details for self-hosted deployments. Use the Claude Code integration to automatically delegate Sentry-related queries to the sentry-mcp server, which will translate natural language requests into Sentry's query syntax and return structured results. If you operate a self-hosted Sentry deployment, you can adjust the host and disable unsupported skills (e.g., Seer) via the provided options. When enabling AI-powered search features, you must supply a configured LLM provider (OpenAI or Anthropic) and their respective API keys.

Prerequisites:

• Node.js installed (for npx-based invocation) or an environment that can run npm/npx commands
• Access to the Sentry MCP server package (@sentry/mcp-server) on the npm registry

Installation steps:

1. Ensure you have an access token for Sentry API if using the standard (non-self-hosted) flow:
   • SENTRY_ACCESS_TOKEN will be required in environment variables
2. Create the MCP config snippet (see mcp_config below) to run the server via npx
3. Install and run the MCP server using the provided command, optionally customizing host and skills:
   • For example, run: npx @sentry/mcp-server
4. If you are self-hosting Sentry, set SENTRY_HOST in the environment to point to your instance and optionally disable unsupported skills via MCP_DISABLE_SKILLS

Additional setup for local development:

• Install dependencies and set up OAuth apps in Sentry as described in the repository docs
• Provide OPENAI_API_KEY or ANTHROPIC_API_KEY depending on your chosen LLM provider
• Start the development server and test via the MCP Inspector or CLI tools as described in the README

Tips and known considerations:

• Always explicitly set EMBEDDED_AGENT_PROVIDER to avoid auto-detection issues; choose either 'openai' or 'anthropic'.
• If your Sentry instance is self-hosted and lacks certain features (e.g., Seer), use MCP_DISABLE_SKILLS to hide those tools.
• For stdio-based testing or self-hosted setups, ensure SENTRY_HOST is correctly configured and that the access token has the necessary scopes.
• The MCP Inspector can be used to verify connectivity and authentication flows; if OAuth redirects fail locally, try using localhost as the host in the MCP URL.
• When using npx in production, pin the version (e.g., @sentry/mcp-server@latest or a specific tag) to avoid unexpected breaks.
• Ensure your environment variables are kept secure and not committed to version control.