You know ripgrep — the benchmark for finding things in files: fast, precise, lightweight.
Red Hat Emerging Tech just applied that same idea to AI coding. It's called ripwire (2024 stars, Apache-2.0, C++23): the ripgrep of AI context — give your coding agent a map before it reads the repo.
The problem it solves
The biggest waste in coding agents is blind reading. Ask an agent to fix a bug and it greps around, slurps whole related files into context, burns tokens, and often reads the wrong places.
ripwire inverts it: before the agent reads, give it a ranked, deterministic call graph — what to touch, what it breaks, which tests to run — instead of letting it grep and guess.
Three signals it's real
- Built by Red Hat Emerging Tech. Not a hobby project — a frontier team with the engineering rigor and maintenance stamina that individual side projects rarely have.
- Zero-dependency C++23. A CLI + MCP server with no runtime dependencies: one command to install, no dependency hell, instant startup, tiny footprint.
- Sourced engineering wisdom. The README states it folds in 49 repositories and 70 papers — from McCabe's 1976 cyclomatic complexity through seven papers published in just the last two months. Every row in its lineage doc names the lesson taken and the file it lives in. That's a different species from "a tool someone knocked together."
And a hardcore detail: its function signatures are 74.7% fewer bytes than their bodies. That translates directly to fewer tokens and faster context — i.e., cheaper and quicker agents.
The honest caveat
It's infrastructure for coding agents, not a tool for humans. You won't open ripwire to read code yourself — you start a Codex or Claude Code session with "use ripwire on this repo" and let the agent use it to understand structure. And it's tuned for C++/systems code (unsurprising, given the authors); front-end or scripting projects may see less benefit.
I've localized the README and core docs to Chinese: https://github.com/yangshun2005/ripwire-cn
If you find this project useful, a star on the original repo supports the author's ongoing maintenance.
Top comments (0)