This is a Model Context Protocol (MCP) server written in Python that acts as a bridge between AI assistants (like Claude Desktop, Cursor, or Zed) and the Sentry.io error tracking platform. It exposes Sentry issue data as tools and prompts that language models can call during conversations.

The problem it addresses is friction in debugging workflows. Developers using AI coding assistants typically have to manually copy error messages, stacktraces, and issue metadata from Sentry into their chat sessions. This server automates that step by allowing the AI to directly query Sentry using an issue ID, URL, or project slug and receive formatted issue details including title, status, level, timestamps, event counts, and full stacktraces.

The primary users are developers who use Sentry for error monitoring and interact with MCP-compatible AI assistants. Configuration is done via CLI arguments (auth token, project slug, organization slug) either through uvx, pip, or Docker, and referenced in the client's MCP config file (e.g., claude_desktop_config.json or .cursor/mcp.json).

Under the hood, the server is small and straightforward. It uses the official `mcp` Python SDK to run over stdio, `httpx` for HTTP calls to the Sentry API at https://sentry.io/api/0/, and `click` for CLI parsing. It exposes two tools (`get_sentry_issue`, `get_list_issues`) and one prompt (`sentry-issue`). The `extract_issue_id` function validates whether input is a numeric ID or a *.sentry.io URL, and `create_stacktrace` formats exception entries from the latest event into a readable trace.

This project is explicitly a fork of the official modelcontextprotocol/servers Sentry implementation, with added packaging for PyPI, Smithery, and Docker distribution. No suspicious patterns were observed — auth tokens are passed as CLI arguments and used only against sentry.io endpoints. The code is minimal (~13KB of Python) and behavior matches the README.