AI-Researcher is an open-source autonomous research agent from the Data Intelligence Lab at the University of Hong Kong (HKUDS). Presented at NeurIPS 2025, it aims to automate the scientific innovation pipeline, from literature review and idea generation through experiment implementation to paper writing. The repository has collected more than 5,800 stars on GitHub, and the team also offers a production-ready hosted version of the system.
AI-Researcher orchestrates large-language-model agents across the whole research workflow. Given a research direction, it surveys existing papers, proposes and refines hypotheses, writes and runs experiment code, analyses the results and drafts a manuscript. The repository includes the agent pipeline together with collected resources and benchmarks for evaluating research agents, making it a useful reference implementation for anyone studying how far end-to-end research automation can currently go. Its authors report experiments in which the system carried a topic from initial survey to a complete draft paper, and the lab continues the line of work with a hosted, production-ready version aimed at working researchers. Even for teams that never run it end to end, the individual stages, such as automated literature surveys and experiment planning, are instructive building blocks. The codebase is organised so those stages can be studied or reused independently.
Free. The research code is published openly on GitHub for anyone to study and build on; a separate production-ready hosted edition is offered through the team’s Novix platform.
AI researchers, graduate students and R&D teams exploring autonomous scientific discovery, or looking for a working blueprint of a full research-agent pipeline.