Curie

Curie is an open source AI agent framework for automated, rigorous scientific experimentation. Given a research question, it works through the experimental process end to end: formulating hypotheses, designing and running experiments, analyzing results, and documenting the outcome in a report with figures. Its emphasis is on reliability and reproducibility rather than open ended generation. It is aimed at researchers and machine learning engineers who want to automate experiment loops. Built in verification modules enforce methodical procedure and check agent reliability at each stage. The framework supports machine learning engineering, systems analysis and general scientific discovery, and it can be customized with a user's own starter code and arbitrary datasets. The team behind it also offers a companion tool, Are you Scooped?, that analyzes a research idea to find whether similar work already exists. Curie is free to use and released under the Apache 2.0 license on GitHub. There is no published paid plan; users supply their own compute and model API access. The project originates in academic research. Its authors published the paper Curie: Toward Rigorous and Automated Scientific Experimentation with AI Agents in February 2025, followed by EXP Bench, a benchmark asking whether AI can conduct AI research experiments, in June 2025. The public repository has seen little activity since late September 2025, so buyers should check its current maintenance status before relying on it.