CodeScene's MCP server brings Code Health analysis directly into your AI assistant workflow. It exposes tools for reviewing code maintainability, identifying technical debt hotspots, calculating refactoring business cases, and validating that AI-generated changes don't introduce quality regressions. The server runs locally and connects to your CodeScene account (Cloud or on-prem) via REST API to pull hotspots and ownership data, while performing all code analysis on your machine. Reach for this when you want AI assistants like Claude or Cursor to make refactoring decisions based on concrete metrics like complexity, cohesion, and Code Health scores rather than just syntax patterns. It includes reusable agent skills and works best with frontier models that can act on nuanced quality signals.
claude mcp add --transport stdio codescene-oss-codescene-mcp-server uvx codescene-mcp-server