🔥 Hot Repo: 120x Token Savings, Zero Runtime Dependencies

One-liner — codebase-memory-mcp is a zero-dependency MCP server that maps any codebase into a persistent knowledge graph, slashing LLM token usage by up to 120x compared to file-by-file code exploration. - Repo: DeusData/codebase-memory-mcphttps://github.com/DeusData/codebase-memory-mcp - Stars: ⭐ 4,273 +367 today - Language: C - License: MIT --- What It Does codebase-memory-mcp is an MCP server written in pure C that parses your entire codebase using tree-sitter AST analysis, storing functions, classes, call chains, and HTTP routes in a persistent SQLite knowledge graph. Instead of an AI agent reading dozens of files to answer "where is this function called?", it runs a sub-millisecond graph query. The result: 5 structural queries cost 3,400 tokens versus 412,000 tokens via traditional file-by-file search — a 120x reduction. Why It's Blowing Up A preprint on arXiv 2603.27277 dropped alongside the v0.8.1 release and validated the approach across 31 real-world repositories: 83% answer quality, 10x fewer tokens, and 2.1x fewer tool calls versus standard file exploration. Developers working with Claude Code and Codex have been hunting for exactly this kind of structural shortcut as context windows fill up and per-token costs compound. The tool also hit a tipping point in breadth: the install command now auto-detects and configures 11 coding agents simultaneously — Claude Code, Codex CLI, Gemini CLI, Zed, OpenCode, Aider, KiloCode, VS Code, OpenClaw, Antigravity, and Kiro. One command, all agents wired up. That kind of zero-friction story spreads fast in developer communities. A secondary driver is the delivery format: a single static C binary with zero runtime dependencies. No Node.js, no Python, no Docker. That eliminates the biggest friction point in MCP adoption — the setup tax that kills adoption before you even try a tool. Key Features - 158-language support — all grammars compiled into the binary via vendored tree-sitter; nothing extra to install -