JCodeMunch

•
2.9k
•

Optimizes AI agent interaction with codebases by enabling structured retrieval and significantly reducing token usage through precise AST parsing.

JCodeMunch acts as a powerful MCP server designed to revolutionize how AI agents explore and understand code repositories. By indexing a codebase once using tree-sitter AST parsing, it transforms brute-force file scanning into precise, symbol-level retrieval. This dramatically cuts code-reading token usage by 95% or more, allowing agents to fetch exact functions, classes, methods, or contextual bundles on demand, rather than wasting resources skimming entire files. It provides structured insights and queries that go beyond traditional text search, fostering faster, more accurate, and cost-efficient AI-driven development workflows.

主な機能

01Includes BM25-based symbol search, intelligent context bundling, and query suggestions
02Ensures rapid access through local-first indexing and live watch-based reindexing
031,214 GitHub stars
04Provides symbol-level retrieval (functions, classes, methods) with byte-level precision
05Achieves 95%+ lower code-reading token usage for AI agents
06Offers advanced structural queries like blast radius analysis and class hierarchy traversal

ユースケース

01Enhancing agent-driven code exploration in large or unfamiliar repositories
02Streamlining refactoring, debugging, and impact analysis workflows
03Reducing AI token costs for development teams and individual developers