Deep Code Reasoning

This MCP server enables intelligent routing between Claude Code and Google's Gemini AI for complementary code analysis, treating LLMs as heterogeneous microservices where each handles tasks suited to their strengths. Built by Jonathan Haas using TypeScript with Google's Generative AI SDK, it provides both traditional single-shot analysis tools and conversational AI-to-AI dialogue capabilities that allow Claude and Gemini to engage in multi-turn problem-solving sessions. The implementation leverages Gemini 2.5 Pro Preview's 1M token context window for analyzing large codebases, distributed system traces, and cross-service impact analysis that exceeds Claude's context limits, while Claude handles local-context operations and incremental patches. It includes security features like prompt injection protection, path traversal prevention, and input sanitization, making it valuable for debugging complex distributed systems, performance regression hunting, and hypothesis-driven analysis where extensive code execution and correlation across multiple services is required.

Source

Repository: https://github.com/evalops/deep-code-reasoning-mcp