AI product Open source
NeuroArxiv is an agent skill that makes Claude Code or Codex CLI check arXiv prior art before proposing non-trivial architectures, algorithms, or systems techniques. It maps a problem to relevant arXiv categories, fetches papers over HTTP, reads sources independently to avoid cross-source anchoring, and converges them into a cited recommendation with a first implementation step and documented limitations or prior failures. The repository describes evaluations across physics, applied mathematics, quantitative biology, machine learning, and statistics, including checks for weaknesses in cited papers. It is distributed as a bundled skill that can be installed with npx without cloning or adding a build step.
1 use taken from transcripts — each links to the moment in the video.
A coding-agent tool that checks arXiv before an agent invents an architecture. It maps a problem to categories, fetches relevant papers, reads abstracts independently, and produces a recommendation with a first step and documented prior failures.
1 in the library.