STORM

STORM is an MIT-licensed, open-source research tool from Stanford OVAL that turns a topic into a Wikipedia-style report with citations. You can run its Python software on your own machine or host it on a server; Stanford also offers a hosted research preview. It's aimed at people gathering background material and writers preparing an article draft.

The research process explores different perspectives on a topic, collects references, and builds an outline before writing the report. That makes its question generation a useful distinction: it draws on related articles to find angles to investigate, rather than relying on a single prompt for a finished answer. Reports still need fact-checking and editing before publication.

Co-STORM adds an interactive discussion with AI experts and a moderator. You can follow their exchange or contribute to steer the research, while a changing mind map organizes the information into related concepts.

For self-hosted use, STORM supports language and embedding models through LiteLLM, including workflows using OpenAI models. It can research through services such as Bing, Brave, Tavily, and SearXNG, or ground its writing in documents you provide through VectorRM. Its research, outlining, writing, and polishing components are customizable.

Self-hosting runs the research software under your control, but internet search and cloud model services still involve external providers. The hosted preview requires sign-in and collects inputs and feedback for research, with possible redistribution under CC-BY or a similar license.