Biomni Lab

by Christian Rudolph

Most AI tools that researchers encounter operate through a single chat window. You type a question, you get an answer. This is useful, and increasingly familiar. Biomni Lab, developed by Phylo, is a slightly different proposition – and one that is worth being aware of if you work in Biomedical Sciences.

Launched in February 2026 as what Phylo describes as the first Integrated Biology Environment, Biomni Lab is designed not as a plain conversational assistant but as a workspace in which AI agents carry out real biological research tasks. You still interact through a chat, but it sits within an environment of several linked panels rather than a single scrolling conversation: populated windows that can be clicked to show plans, results, code and analyses side by side. Building on an open-source project that began at Stanford, the platform integrates hundreds of biological tools, databases and analytical systems into that single environment – around sixty curated databases among a much larger set of software packages and specialised tools. It allows scientists to orchestrate agents to plan, author, execute and synthesise across them.

What I used it for

I tried Biomni Lab for a specific and reasonably demanding task: a detailed literature analysis on a problem in bacterial DNA physiology. The question was narrow, technical and required engagement with a body of primary literature where the details matter – the kind of analysis where a superficial or pattern-matched response would be immediately obvious.

The result was extensive and, more importantly, technically careful. The analysis engaged with the right primary literature, drew the relevant mechanistic distinctions and presented the material at a level of specificity that reflected genuine engagement with the sources rather than a generic summary. It was, in short, the kind of output that saves real time – not because it replaced thinking, but because it did the groundwork that thinking then builds on.

The learning curve

I should be honest about one thing: Biomni Lab is not immediately intuitive. If you arrive expecting a single chat window, you will need to adjust. The interface is more complex, the workflow less obvious, and there is a period of familiarisation before the platform starts to feel natural. This is not a criticism so much as a warning: the additional complexity exists because the platform is attempting something considerably more ambitious than a standard AI assistant, and that ambition has a cost in accessibility.

The potential, however, is clear. I have only used one feature of what is evidently a much larger environment. The platform promises capabilities well beyond literature analysis – experiment planning, bioinformatics pipelines, model fine-tuning – none of which I have yet explored. What I can say is that the one thing I did try was impressive enough to make further exploration worthwhile.

Worth trying

If you work in biology and regularly need to get to grips with a body of literature quickly and accurately, Biomni Lab may just be worth the investment for literature work alone, with much more beyond it. It is currently available in research preview.

You can find it at biomni.phylo.bio.

I came across Biomni Lab because Stuart kindly sent me a link to it. I have the feeling I might turn into a regular user.

Similar and related blog posts:

• Peer review of a manuscript using only a prompt
• Peer review of a manuscript by the q.e.d. platform
• Peer review of a manuscript by LINER AI
• Peer review of a manuscript by Paper Wizard