The Sentry MCP supports both **local stdio** and **remote** connections: - **Authentication**: Both modes require Sentry authentication—OAuth for remote, API token for local - **Organization Access**: Access is scoped to organizations and projects you have permissions for - **Sensitive Data**: Error reports may contain sensitive information like user data or environment variables - **Remote Option**: The remote server (`https://mcp.sentry.dev/mcp`) runs on Sentry's infrastructure **Security best practices**: Use organization-scoped API tokens with minimal permissions, be cautious when sharing error details that may contain sensitive data, and review which projects the integration can access.

You can connect via local stdio or remote server: **Option 1 - Remote (Recommended)**: 1. Configure your AI client to connect to `https://mcp.sentry.dev/mcp` 2. Complete the OAuth flow to authorize access **Option 2 - Local**: 1. Generate a Sentry API token from your account settings 2. Configure your AI client with `npx @sentry/mcp-server` 3. Set your `SENTRY_AUTH_TOKEN` environment variable **Example prompts**: - "What are the most frequent errors in my project this week?" - "Analyze issue PROJ-123 and suggest a fix" - "Find all errors in the authentication module"