CMIP-Forge: An Agentic System that Retrieves, Computes, and Self-Reviews Climate Science

CMIP-Forge is a hybrid retrieval-augmented generation (RAG) and autonomous analysis system designed to bridge the gap between climate science literature and live data analysis, utilizing a curated corpus of 6,581 CMIP6-related publications. It features a multi-layered Defense-in-Depth architecture that enforces methodological invariants through Abstract Syntax Tree (AST) analysis and an autonomous adversarial peer-review protocol, enabling end-to-end autonomous research workflows in climate science. This system addresses critical bottlenecks in operationalizing climate knowledge, offering practitioners a robust framework for conducting complex analyses while ensuring methodological rigor through automated auditing.