repobrain

Why we included this project

Developers who live inside AI coding agents know the ritual: every session starts by re-explaining what the repository does before the model can be useful. RepoBrain removes that by scanning the workspace once, building a knowledge doc for each module, then answering questions with file paths and line numbers attached so every response can be checked against the code. That grounding matters because agents tend to guess at architecture when they only see a large tree piecemeal. The index is stored in a plain .repobrain/ folder in the project, so the same knowledge base follows you into Claude Code, Cursor, Codex, Windsurf, or other hosts, which helps small teams switching editors or onboarding people into unfamiliar codebases. If the recurring frustration is asking a model how the project.