@Sumanth_077 explains jevgrep: Jev judges which folders, files, and declarations are relevant, then returns source excerpts for the coding agent.

Guide to someone else's open-source CLI. The post says jevgrep is for when you know a behavior but not where it lives. If you already know the symbol, filename, or string, grep, rg, or file search is the better fit. An agent skill can call jg from Claude Code, Codex, and OpenCode. On a 10-task SWE-bench setup they cite, both jevgrep and the no-Jev baseline solved 8/10, and coding-agent cost fell from $7.62 to $5.44, excluding Jev cost. The GitHub link is in the comments, not the body. Includes a photo and quotes another post. Verified via fxtwitter.