FutureHouse opens its scientific agent platform: Crow, Falcon, Owl and Phoenix reachable through a web interface and API
Synopsis
FutureHouse has launched a publicly available platform of scientific agents, with four agents covering general literature question answering, deep literature review, a check on whether anyone has done a given thing before, and chemistry experiment planning, reachable through a web interface and an API. The platform says Crow, Falcon and Owl outperform all major frontier search models on retrieval precision and accuracy, and beat PhD-level researchers in head-to-head literature search tasks on precision; Phoenix is labeled experimental, less thoroughly benchmarked than the other three, and may make more mistakes.
Interpretation
Researchers can now call four literature-focused scientific agents directly through a web interface and an API instead of building and maintaining their own agent deployments. The platform offers this kind of scientific agent publicly to all researchers, while self-maintained agent deployments are very difficult for scientists to keep running. The platform launches with four agents, Crow, Falcon, Owl and Phoenix, with Phoenix labeled experimental and described as less thoroughly benchmarked and more error-prone than the others.
The agents are built around scientific literature retrieval and synthesis, with access to a large corpus of high-quality open-access papers and specialized scientific databases such as OpenTargets. The platform says that, unlike other agents, FutureHouse agents can reach specialist scientific databases and automate chemistry workflows. Crow, Falcon and Owl have been benchmarked, and the platform says they outperform all major frontier search models on retrieval precision and accuracy and exceed PhD-level researchers in precision in head-to-head literature search tasks.
The agents use a multi-stage process to examine each source in more depth and expose that reasoning so users can see how a conclusion was reached. The platform says that, unlike other agents, FutureHouse agents have transparent reasoning and consider each source in more depth. The platform states that every user can view this process to tell exactly how the agents arrived at a given conclusion.
Perspective
The platform targets researchers who need full-text literature analysis and specialized chemistry tools, and is offered through a web interface and an API; Phoenix is labeled experimental, described as less thoroughly benchmarked and more error-prone than the other agents, and released in the spirit of rapid iteration.
The platform says Crow, Falcon and Owl outperform all major frontier search models on retrieval precision and accuracy and exceed PhD-level researchers in precision in head-to-head comparisons, but gives no specific figures; Phoenix is less thoroughly benchmarked and may make more mistakes.
