’ve built NeuroCode, a Python engine that builds a structural intermediate representation (IR) of codebases — including call graphs, module dependencies, and control flow — and stores it in a neural-ready format designed for LLMs.
Most tools treat code as text. NeuroCode treats it as structure. It gives you a CLI (and library) to:
Build the IR with neurocode ir . and store it as .neurocode/ir.toon
Explain files using call/import graphs: neurocode explain path/to/file.py
Run structural checks and generate LLM-ready patch plans (coming soon)
The goal is to bridge static analysis and AI reasoning. You can plug NeuroCode into agents, editors, or pipelines — or use it standalone to get structure-aware insights into your codebase.
No runtime deps, tested with Python 3.10–3.12. Still early (v0.3.0), feedback and contributions welcome.
GitHub: https://github.com/gabrielekarra/neurocode