BioRender unveils Leo, an AI agent for building scientific figures

BioRender has unveiled Leo, which the company calls the first AI agent built specifically for making scientific figures. The launch, announced in a Business Wire release, targets a narrow but concrete problem: helping researchers turn ideas into publication-ready diagrams.

Unlike generic AI image tools that return a single flat picture, Leo works through conversation. It asks focused questions about your key message, audience, components and the relationships between them, then produces a sketch grounded in your science.

From the sketch, researchers can mark up the draft — crossing out a component, drawing in a missing step, or circling a finding to emphasize. Leo turns that markup into a polished figure, which BioRender says takes a fraction of the usual time.

Editability is the other differentiator. The finished figure opens on the BioRender canvas, where every label, arrow, icon and shape stays a separate object that can be changed individually, or revised by asking Leo to adjust one part.

BioRender positions Leo against generic AI image tools, which it says return a single flat picture that must be regenerated for any change. In its comparison, Leo supports planning a figure's message and structure in conversation, marking up a sketch before the full figure is built, keeping a consistent aligned layout, and editing any element without regenerating.

On data use, BioRender says your prompts, conversations, uploads and figures are not used to train its own or its AI providers' models without consent. It also cautions that Leo offers a strong starting point but does not guarantee scientific accuracy, so researchers should review the output before use.

For scientific workflows, the significance is a shift from a one-shot generation to an iterative, conversational process that keeps every element editable. What to watch next is how Leo holds up in real papers and grant applications, and whether it spurs more purpose-built agents for scientific imagery and data.