FutureHouse is launching our platform, bringing the first publicly available superintelligent scientific agents to scientists everywhere via a web interface and API. Try it out for free at https://platform.futurehouse.org.
The new FutureHouse Platform
Science is bottlenecked by data. The 38 million papers on PubMed, 500,000+ clinical trials, and thousands of specialized tools have created an information bottleneck that even the most brilliant scientists can’t navigate. At FutureHouse, our mission is to solve this problem by building an AI Scientist. Today, we are taking a significant step forward by releasing the first publicly available superintelligent scientific agents accessible to researchers everywhere, with benchmarked superhuman literature search & synthesis capabilities.
The FutureHouse Platform is launching with four agents, each with their own specialization:
Crow is a general-purpose agent that can search the literature and provide concise, scholarly answers to questions, and is perfect for use via API.
Falcon is specialized for deep literature reviews. It can search and synthesize more scientific literature than any other agent we are aware of, and also has access to several specialized scientific databases, like OpenTargets.
Owl (formerly HasAnyone) is specialized to answer the question “Has anyone done X before?”
Phoenix (experimental) is our deployment of ChemCrow, an agent with access to specialized tools that allow it to help researchers in planning chemistry experiments.
Each agent is built from the ground up for science. Crow, Falcon, and Owl have been rigorously benchmarked [1, 2, 3], and outperform all the major frontier search models on retrieval precision and accuracy. We have also experimentally validated their retrieval and synthesis abilities as having better precision than PhD-level researchers in head to head literature search tasks.
Phoenix is not as deeply benchmarked as our other agents and may make more mistakes. We are releasing it now in the spirit of rapid iteration. Please provide feedback as you use it!
Several factors set our agents apart
Unlike other agents, FutureHouse agents have access to a vast corpus of high-quality open-access papers and specialized scientific tools, allowing them to automate workflows in chemistry and to retrieve information from specialist scientific databases. They can also use a variety of methods to evaluate source quality, just like a researcher would.
They have transparent reasoning and use a multi-stage process to consider each source in more depth. Moreover, every user can see this process in order to tell exactly how the agents arrived at a given conclusion.
Finally, our platform is built for scale. It is very difficult for scientists to maintain their own agent deployments, so we are providing an API, in addition to a web interface, to facilitate researcher workflows.
By chaining these agents together, at scale, scientists can greatly accelerate the pace of scientific discovery.
The platform is particularly effective for challenges requiring detailed, full-text literature analysis, or (using Phoenix) challenges that require use of specialized chemistry tools. Here are some examples of how you can use it:
- Identify unexplored mechanisms in disease pathways: As outlined in the video, scientists can use Falcon for background knowledge, Crow to identify key genetic associations, and Owl to determine where the research gaps exist - all in minutes rather than weeks of literature review.
- Systematically identify contradictions in the literature: Scientists can task Falcon with analyzing conflicting evidence across hundreds of papers on controversial topics and highlight where additional experiments could resolve these conflicts.
- Conduct critical analyses of the methods used to obtain a given result: Because our agents have access to full text papers, you can ask them questions about experimental methods or study limitations that might not otherwise be apparent from the abstract.
- Customize research pipelines via API: Research groups can build automated systems that continuously monitor new publications, or that conduct literature searches at scale to contextualize the results of screening experiments.
- Find known hits that bind to a target protein: Scientists can task Phoenix to use data from existing sources to propose hits to a protein target along with complex constraints on solubility, functional groups, or novelty.
- Reason about chemical space: Phoenix can tell you if compounds are novel, how much they cost, and predict the outcome of a reaction. It can even tell you if it's cheaper to buy or make a compound.
Try the platform out for yourself at https://platform.futurehouse.org. We are extremely excited to see how researchers everywhere are able to use our agents to supercharge their work — if you have use cases you'd be excited to share with us, tag us on X or email us at admin@futurehouse.org!
