- Rust
Overview
Designed with both agentic and human workflows in mind. Goal is deterministic review tools that don’t consume tokens, offloading work from agents via skills/mcp tooling.
Why?
Working on large codebases can be a pain. Navigating code, changing code with hidden dependencies that break critical functionality, lacking documentation and test coverage, AI code reviews performing poorly due to overwhelmed context windows, etc. I wanted a tool that allows users to explore codebases and analyze them for high-risk/highly-dependent areas.
There are many great open-source packages for this that do some, or even all of this capability already. I wanted to learn Rust and needed something to build, so I figured why not make my own.
I also had the idea: what if we can natively support skills/MCP so that agents can use this tool for initial review step? Hopefully reducing token spend and context window usage of basic and inherently deterministic/static properties of codebases.
Is this actually useful? No idea. Eventually, I plan to develop benchmarks and testing capabilities to compare agent performance with/without codegraide + comparisons against other OSS. But as with all things currently, post-training optimization will most likely to continue to reduce this tool’s benefits lower and lower.