jgravelle/jcodemunch-mcp

Symbol-level code retrieval for AI agents: fetch the function you need, not the whole file. Runs locally. 96.2% fewer tokens per lookup than grep-and-read on our 15-lookup benchmark. Works with Claude Code, Cursor, Codex and any MCP client.

What it solves

AI agents often waste tokens by reading entire files to find specific pieces of code, which increases costs and consumes the context window. jCodeMunch provides a way for agents to retrieve only the exact code they need—such as specific functions, classes, or methods—with byte-level precision, significantly reducing token usage.

How it works

The project indexes a codebase once using tree-sitter AST parsing to create a local index of structured symbol metadata (signatures, kinds, and byte offsets). Instead of reading whole files, agents can query this index to find symbols and fetch only the exact implementation of a specific symbol on demand.

Who it’s for

Developers using AI agents (like Claude Code, Cursor, Windsurf, or Continue) who work with large repositories or unfamiliar codebases and want to reduce AI spend and improve agent reliability.

Highlights

  • Extreme Token Efficiency: Benchmarked at 28.3x fewer tokens than a grep-and-read approach, reducing costs by 86-99%.
  • Structural Queries: Enables agents to perform complex analysis that grep cannot, such as finding importers, calculating "blast radius" for changes, and detecting dead code.
  • Broad Language Support: Supports over 70 languages via tree-sitter.
  • Local-first Design: Indexes are stored locally, and the system includes automatic secret redaction before code is sent to the LLM.
  • MCP Compatible: Works as a Model Context Protocol (MCP) server, integrating with a wide variety of AI-enabled IDEs and CLI tools.

Related

  • Project
    johnhuang316/code-index-mcp

    An MCP server that provides intelligent code indexing and AST-based analysis to help AI assistants navigate and understand complex codebases.

  • Project
    DeusData/codebase-memory-mcp
  • Project
    Mibayy/token-savior

    An MCP server that reduces AI coding agent token usage and costs by providing structural code navigation, persistent memory, and Bash command rewriting.

  • Project
    forloopcodes/contextplus

    An MCP server that turns large codebases into searchable, hierarchical feature graphs using RAG, AST parsing, and spectral clustering for high-accuracy AI coding context.

  • Project
    aovestdipaperino/tokensave

    tokensave is a Rust‑based local server that builds a semantic knowledge graph of a codebase, letting AI coding agents query symbols, relationships, and snippets in a single call. By replacing costly grep/glob scans, it cuts token usage, supports multi‑branch indexing, offers crash‑resilient syncing, and integrates with dozens of LLM‑powered coding tools.