Deep codebase explorer that explains why code works, with on-demand doc export
How it works
The user provides a repo location (a GitHub URL or a local path). The agent indexes the codebase using a tree-sitter-style semantic parser so it understands structure, not just text.
The user can then ask free-form questions in chat: high-level ("what is the architecture of this project?"), module-level ("how does the auth flow work and why is it done this way?"), function-level ("walk me through this function and its edge cases"), or line-level ("what does this exact line do?"). The agent picks the right granularity based on the question, returns a clear inline answer in the chat, and after each answer asks the user if they want the explanation saved as Markdown docs (per-question file or a single growing doc — agent suggests the best shape). If the user says yes, it writes the docs into the repo (or a docs/ folder) and confirms the path. Over time the user builds a navigable knowledge base of "why" answers for the repo.
Example Prompt
“Point this at a repo (GitHub URL or local path) and ask it things like "why does this auth flow work this way?" — get a clear inline answer, then optionally save it as Markdown docs.”