getsentry/toolkit

What it solves

It provides a bridge between Sentry's error monitoring and debugging data and AI coding agents. This allows developers to use AI assistants (like Cursor or Claude Code) to directly query, analyze, and triage Sentry errors, issues, and performance traces without leaving their development environment.

How it works

The project implements a Model Context Protocol (MCP) server that acts as middleware between an AI agent and the Sentry API. It supports two primary transport methods:

• Remote: A hosted service (via Cloudflare) that uses OAuth or direct Sentry tokens for authentication.
• Stdio: A local transport for self-hosted Sentry installations.

For AI-powered search tools (e.g., search_errors), the server integrates with LLM providers (OpenAI, Anthropic, or OpenRouter) to translate natural language queries into Sentry's specific query syntax.

Who it’s for

Developers using AI-powered coding assistants who want to integrate Sentry's debugging and performance data into their automated workflows.

Highlights

• Agent Integration: Specifically optimized for tools like Cursor and Claude Code.
• Natural Language Search: Translates human language into Sentry query syntax using LLMs.
• Flexible Deployment: Supports both a managed remote service and local stdio transport for self-hosted Sentry.
• Skill-based Tooling: Allows users to enable or disable specific "skills" (e.g., inspect, seer, project-management) to control which tools are exposed to the agent.