Math Scout

Math Scout is an experimental multi-agent reading aid for mathematical papers written in LaTeX. It sends sections of a paper to specialised reviewers that look for possible proof gaps, missing hypotheses, brittle arguments, unclear exposition, notation drift, and claims that go beyond the stated results, then synthesises their findings into a single report.

It is intended as a first pass, not as a proof checker or a substitute for human refereeing: it can miss genuine errors and flag correct arguments. Reviews can use Anthropic models, OpenAI models, or both, and sending a paper for review shares its text with the selected API providers.