FutureHouse
AI research agents for scientific literature and discovery
FutureHouse works on automating scientific research. The lab builds language-model agents that read and reason over the research literature, and makes some available to scientists on a platform. Crow answers questions with cited sources from papers, Falcon performs deeper literature reviews, and Owl helps check whether a topic has been studied.
A more ambitious project, Robin, links agents into a multi-step workflow. Agents suggest hypotheses, plan experiments, and analyze results, and the lab used the system to propose a candidate treatment, which was then tested experimentally. The goal is to speed up discovery by letting agents handle parts of the loop.
FutureHouse is a nonprofit, and its tools are free to use, with usage limits. AI research assistants can misread papers, so claims should be verified against sources. Alternatives include Elicit, Consensus, Undermind, Google’s AI co-scientist, and SciSpace. FutureHouse suits scientists exploring AI-assisted literature work.