NVIDIA BioNeMo Agent Toolkit - a collection of scientific computing tools for AI agents - is now available with more than 50 leading companies already adopting it. The toolkit enters a global R&D sector that spends $3.8 trillion annually, targeting biology, chemistry, genomics, and drug discovery to compress timelines that often stretch for years.
What the toolkit contains
BioNeMo Agent Toolkit bundles over a decade of NVIDIA life sciences libraries, open models, and GPU-accelerated software. It relies on NVIDIA NIM microservices, Parabricks for genomics, NeMo, and the Nemotron family of reasoning models. Developers can transform a general-purpose agent into a life sciences agent in minutes, giving it the ability to call domain-specific tools, interpret results, and run multi-step scientific workflows.
"Frontier models are the brains. BioNeMo is the scientific toolbox. Together, they give AI agents the skills of a PhD research assistant and the speed of a supercomputer," said Jensen Huang, founder and CEO of NVIDIA. "For the first time, researchers can build AI agents that understand scientific knowledge, use scientific tools and execute scientific workflows. This is a new way to do science - one that can dramatically accelerate discovery across biology, chemistry, genomics and medicine."
Accelerated workflows from protein design to drug screening
The toolkit turns common lab-tasks into agent-executable routines. In virtual screening, an agent can generate small-molecule candidates, dock them to a protein target, predict binding strength, and rank them by drug-like properties - compressing a process that normally takes days into minutes. For genomic analysis, raw sequencing data can be aligned and variant-called through Parabricks, then scored by genomic foundation models to surface the most disease-relevant targets.
Protein binder design, deep biomedical research, and medical imaging analysis also get agent-ready toolchains. NVIDIA's Biomedical AI-Q Research Agent connects real-world data to reasoning models for literature review, protocol generation, and clinical trial screening. Medical imaging workflows let agents segment and synthesize images to support biomarker discovery.
Collaboration with the University of Washington's Institute for Protein Design has already delivered a 2x speedup for RosettaFold3. David Baker, director of the institute, said: "The next leap in science won't come from a single discovery; it will come from the speed of iterative designs and agents that can repeatedly reason through the complexity of biology at a speed humans never could."
Who is using the toolkit
More than 50 organizations have signed on, spanning frontier AI labs, pharmaceutical firms, and scientific software platforms. Anthropic, OpenAI, Lila Sciences, and Owkin are integrating BioNeMo to help their agents move beyond answering questions and start completing scientific work. Drugmakers like Lilly and diagnostics company Natera are using it to scale repeatable workflows across discovery and clinical insight.
Data and workflow platforms including Databricks, Snowflake, Benchling, and Certara are connecting their systems to BioNeMo-powered AI. Computer-aided drug discovery providers - Dassault Systèmes, Schrödinger, and Cadence (OpenEye) - are embedding the toolkit's capabilities into the applications researchers already use, so that agents can orchestrate molecular generation, docking, and prediction inside existing design platforms. This shift toward agentic science mirrors broader trends in AI Agents & Automation, where AI systems increasingly handle multi-step tasks with minimal human intervention.
Availability
BioNeMo Agent Toolkit and its skills are available now through the NVIDIA developer resources page and GitHub.
Why this matters for researchers
BioNeMo removes a stubborn bottleneck: the manual, repetitive cycles of computational experimentation. Agents can now gather evidence, invoke models, and interpret outputs, allowing scientists to spend more time on hypothesis generation and experimental design. As the cost and time per iteration drop, labs can run more experiments and pivot faster when data contradicts an assumption. The toolkit fits into a growing family of tools for AI for Science & Research, making it easier to deploy AI-driven workflows without building every component from the ground up.
Your membership also unlocks:
