Biomni

Biomni is a general-purpose biomedical AI agent from Stanford that executes research tasks against real tools and databases rather than answering from model memory. It is one of the more credible academic entries in the agentic-science space, and unusually for the category, it is freely available.

What it does

The agent plans a task, selects from a large library of biomedical tools and datasets, executes, and reports what it did. Grounding matters here: the project added GeneCards data specifically so that gene-level queries resolve against a curated source instead of being generated. That design — retrieve and execute rather than recall — is what separates a useful agent from a confident one.

Strengths

• Free and academically developed, with published methodology.
• Executes against real biomedical tools and databases.
• Grounded in curated sources such as GeneCards.
• Broad task coverage rather than a single narrow function.

Limitations

Agentic systems fail in ways that are hard to spot: a plausible plan executed against the wrong dataset produces a confident, wrong answer. Verify the tool calls, not just the conclusion. As an academic project, support and long-term maintenance are less certain than a commercial product. See our coverage of the GeneCards integration.