codebase-memory-mcp

Why we included this project

Coding agents spend a lot of their context budget re-reading files to answer basic structural questions. This MCP server heads that off by indexing the codebase into a persistent knowledge graph: it parses source with tree-sitter across 158 languages and records functions, classes, call chains, HTTP routes, and cross-service links, then exposes the result through 15 MCP tools for search, tracing, impact analysis, and dead-code detection. One graph query can replace a dozen file reads, which is exactly what you want in a large monorepo where the context window fills up fast. It runs as a single static binary with no runtime dependencies and no API keys, so setup is a download-and-install step rather than a deployment project. Teams on Claude Code, Cursor, Codex, or similar agents get reliable structural answers without the code ever leaving the machine.