BioNeMo Agent Toolkit — NVIDIA wants to turn AI agents into lab partners
BioNeMo is here to change the way science works — and that is not an exaggeration.
NVIDIA just launched the BioNeMo Agent Toolkit, a set of tools built specifically for the era of artificial intelligence agents applied to the life sciences. The idea is simple but powerful: bring together more than a decade of NVIDIA libraries, models, and technologies in one place, allowing AI agents to work side by side with scientists and laboratories to accelerate scientific discovery. Think of it this way — instead of a researcher spending weeks reviewing literature, running computational experiments, and cross-referencing data, an agent equipped with this toolkit can do all of that in minutes, with greater precision, and even suggest the most promising next steps. 🚀
Companies like Lilly, Schrödinger, Databricks, Snowflake, and Dassault Systèmes, along with giants like Anthropic and OpenAI, are already adopting or integrating the toolkit into their workflows. More than 50 organizations around the world are already using accelerated tools for tasks ranging from protein design to drug candidate screening. This is a real game-changer for anyone working in scientific research and development. 🧬
What is the BioNeMo Agent Toolkit and why it matters so much
To understand the real impact of this release, it helps to take a step back and look at the problem it solves. Modern scientific research — especially in computational biology, medicinal chemistry, and genomics — involves staggering volumes of data, complex models, and steps that historically depended on large teams, a lot of time, and expensive infrastructure. A single drug discovery cycle, for example, can take years and cost billions of dollars before any concrete results appear. BioNeMo steps right into that gap, adding a layer of artificial intelligence that does not just automate processes but redefines them from scratch.
The toolkit is built on a solid foundation of more than ten years of NVIDIA research and development in biological language models, molecular simulations, and GPU acceleration. It brings together technologies like NVIDIA NIM microservices, NVIDIA Parabricks, NVIDIA NeMo, and NVIDIA Nemotron, along with blueprints like NemoClaw for safe and private agents, and the OpenShell runtime that provides a controlled execution environment. All of this works within an agent-oriented architecture, which means we are not talking about isolated tools. We are talking about systems that can reason through a scientific problem, search databases for information, run simulations, interpret results, and propose hypotheses — all in a chained and autonomous way. It is the concept of a scientific workflow taken to an entirely new level, where AI is not just a passive assistant but an active collaborator inside the lab.
And the best part is that this ecosystem was designed to be modular and integrable. In other words, a research team that already uses platforms like Databricks or Snowflake does not need to abandon its existing infrastructure to tap into the power of BioNeMo. Integration happens naturally, plugging AI agents directly into data pipelines already in use. This drastically lowers the barrier to entry and makes adoption far more accessible for organizations of different sizes and levels of technological maturity.
Jensen Huang captures the spirit of the initiative
NVIDIA founder and CEO Jensen Huang left no doubt about the ambition behind the project. In his words, frontier models are the brain and BioNeMo is the scientific toolbox. Together, they give AI agents the skills of a PhD-level research assistant and the speed of a supercomputer. For the first time, he said, researchers can build AI agents that understand scientific knowledge, use scientific tools, and execute scientific workflows — a new way to do science that can dramatically accelerate discoveries in biology, chemistry, genomics, and medicine.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
That perspective was reinforced by David Baker, professor of biochemistry at the University of Washington and director of the Institute for Protein Design (IPD). Baker emphasized that every tool they build for protein design is only as powerful as the ability of scientists to access it efficiently. The next leap in science, he said, will not come from a single breakthrough but from the speed of iterative designs and agents capable of reasoning repeatedly over the complexity of biology at a speed humans could never match. The collaboration between the IPD and NVIDIA has already delivered 2x faster performance for cutting-edge biodesign models like RosettaFold3, and several additional applications to accelerate protein design are still in progress.
How the scientific workflow changes in practice
When we talk about transforming a scientific workflow, it is easy to get lost in abstractions. So let us be concrete. Imagine a team of researchers working on designing a new molecule with therapeutic potential. Before BioNeMo, that team would need to manually sift through scientific literature, run protein structure prediction models, simulate molecular interactions, and filter results based on predefined criteria — a process that could take weeks just in the exploratory phase. With the toolkit agents, that same flow can be compressed to hours or even minutes, with the agent automatically identifying the most promising candidates and presenting a prioritized list for human review. The researcher gets to focus where it truly matters: strategic decision-making and experimental validation.
The original NVIDIA article details five types of workflows that toolkit agents can execute, and each one is worth knowing:
- Virtual Screening: Agents help researchers identify small-molecule drug candidates by generating and screening compounds, performing molecular docking against a target, predicting binding strength, and filtering by drug-like properties. The agent then highlights which candidates should be prioritized, compressing screening timelines from days to minutes.
- Genomic Analysis and Target Discovery: Agents transform raw sequencing data into prioritized genetic insights and biological targets. NVIDIA Parabricks accelerates alignment and variant calling, while genomic foundation models score variant effects and the agent ranks the most relevant candidates for diseases.
- Protein Binder Design: Agents help researchers design and validate candidates computationally before experimental work begins, compressing design steps that traditionally require extensive manual effort.
- Deep Biomedical Research: Agents connect real-world data to reasoning models to improve efficiency and accuracy across scientific and clinical development processes, including literature review, protocol generation, clinical trial screening, and pharmacovigilance.
- Medical Image Analysis: Agents process, segment, synthesize, and reason over medical imaging data to support biomarker discovery, accelerating evidence generation in research workflows.
Another point worth highlighting is the ability of agents to cross knowledge domains seamlessly. An agent equipped with the BioNeMo Agent Toolkit can, for example, combine genomic data with protein structure information, efficacy history of similar compounds, and even toxicity signals — all in a single integrated analysis. Doing that manually within a reasonable timeframe would be practically impossible, but NVIDIA models were trained precisely for this kind of multidimensional reasoning. Scientific discovery stops being a linear process and becomes a networked one, far richer and more efficient.
The complete ecosystem forming around BioNeMo
The impact of the BioNeMo Agent Toolkit goes well beyond a standalone NVIDIA product. What is taking shape is an entire ecosystem of partners, integrators, and users that covers virtually the entire life sciences value chain. The original NVIDIA article makes this very clear by listing the different categories of organizations involved.
Frontier labs and scientific agent builders like Anthropic, Edison Scientific, Lila Sciences, OpenAI, and Owkin are integrating BioNeMo so their agents move beyond simply answering questions and start completing real scientific work. NVIDIA accelerated models and analysis libraries help shorten the path from hypothesis to insight.
Scientific data and workflow platforms like Benchling, Certara, Databricks, Snowflake, and Seqera are using the toolkit to connect their data systems with AI-powered science. BioNeMo skills let agents query biological and chemical datasets, prepare model-ready inputs, launch reproducible workflows, analyze outputs, and return insights directly within the platforms scientists and data teams already use every day.
Pharmaceutical and diagnostics companies like Lilly and Natera are using the toolkit to scale repeatable agentic workflows in discovery, translational research, and clinical insights. Lilly, one of the largest pharmaceutical companies in the world, has been integrating artificial intelligence solutions into its R&D pipelines for several years, but the arrival of a platform like BioNeMo represents a significant qualitative leap. Instead of using AI models in a piecemeal and isolated fashion, the idea is to build complete workflows where agents take over entire stages of the research process, freeing scientists to focus on the decisions that truly require human expertise.
AI-native biology companies like Boltz, Basecamp Research, Chai Discovery, Dyno, PerturbAI, and Proxima collaborated with NVIDIA to develop tools that accelerate model-powered therapeutic design workflows.
Computer-aided drug discovery software providers like Dassault Systèmes, Cadence (OpenEye), and Schrödinger are integrating toolkit capabilities into scientific applications used by discovery teams. This way, agents can orchestrate molecular generation, docking, and prediction, transforming computer-aided design platforms into systems where researchers can ask questions, launch analyses, and identify the best next steps faster.
Lab instrumentation and automation companies like Automata, HighRes, Tecan, Thermo Fisher, and the autonomous data generation platform Medra are connecting their systems with computational discovery powered by BioNeMo skills.
Cloud and AI infrastructure companies like Baseten, Modal, and Nebius are using the toolkit to help developers build life sciences workflows as reliable hosted services, moving agentic biology workflows from prototypes to accessible services for researchers and companies.
Open research organizations are in the game too
It is not just commercial companies jumping on board. Open model and research organizations, including the Arc Institute, the Open Molecular Software Foundation, and the University of Washington Institute for Protein Design, are working with NVIDIA to use BioNeMo in advancing frontier models and making them more accessible through agent-ready workflows. This gives researchers tools at a scale and cost that were not previously possible — democratizing access to computational capabilities that until recently were limited to large corporations with multimillion-dollar budgets.
Numbers that put this technology’s potential in context
To size up the opportunity, it is worth looking at the sector numbers. Global investment in scientific R&D has already reached $3.8 trillion, with annual budgets in the pharmaceutical sector alone approaching $300 billion. Life sciences represent one of the most important scientific frontiers in the world, and any efficiency gain in this space translates directly into faster delivery of new treatments, therapies, and diagnostics to people. BioNeMo agentic workflows can help the industry iterate faster, reduce costs, and maximize the probability of success in every research project.
Accelerated tools already transforming real organizations
The number of more than 50 organizations adopting BioNeMo accelerated tools is not just an impressive metric — it is a clear indicator that the market already recognizes the concrete value of this technology. Among the most relevant use cases are protein structure prediction, molecular docking, generative chemistry, genomic analysis, protein design, and biomarker discovery. With NVIDIA toolkit, these capabilities gain an intelligent orchestration layer that makes them even more powerful and accessible.
A fundamental difference the toolkit brings is the ability to turn general-purpose agents into life sciences specialists in a matter of minutes. A generic agent may struggle to navigate scientific workflows efficiently, needing to infer the right tools, inputs, outputs, and biological meaning along the way. With the BioNeMo Agent Toolkit, agents can call the right tools, interpret results more accurately, and arrive at scientific insights faster and more reliably. This is possible because NVIDIA is optimizing the entire BioNeMo platform, turning libraries, models, and frameworks into tools that agents can call directly.
It is not just the pharmaceutical sector that benefits. The architecture of the BioNeMo Agent Toolkit was designed to be applicable in any area that involves biological data at scale — from agroscience and biotech to basic research at universities and public institutes. The fact that players like Anthropic and OpenAI are also part of the ecosystem suggests the integration potential goes beyond life sciences, opening the door for interdisciplinary collaborations that could generate unexpected, high-impact discoveries. 🌱
What to expect from the future with AI at the center of scientific research
The trajectory of BioNeMo points to a future where the pace of scientific discovery will be increasingly driven by computational capacity and the quality of available artificial intelligence models. That does not mean human scientists will become obsolete — quite the opposite. What changes is their role within the process. Where a large share of time used to be consumed by repetitive, operationally heavy tasks, that time can now be redirected to what the human mind does best: questioning assumptions, asking new questions, and interpreting results within a broader context. Artificial intelligence takes on the heavy lifting, and the scientist steps into the strategic role.
As models evolve and agent architectures mature, we will likely see labs completely redesigned around hybrid workflows in the coming years, where humans and AI agents collaborate in real time. The scientific workflow of the future will be more agile, more connected, and far more data-driven than anything we have today. Organizations that start building this infrastructure now — integrating tools like the BioNeMo Agent Toolkit into their processes — will hold a significant competitive advantage when this landscape fully takes shape.
BioNeMo Agent Toolkit availability
The BioNeMo Agent Toolkit and its skills are already available through NVIDIA developer resources page and the GitHub repository. That means any research team or company interested in exploring AI agent capabilities for life sciences can start experimenting right now, without waiting for early access programs or waitlists.
And for those still watching from the sidelines, it is worth keeping an eye on NVIDIA next moves in this space. The company has been consistent in expanding its scientific AI capabilities, and the toolkit launch is clearly another step in a long-term strategy. The ecosystem around BioNeMo is already robust, with heavyweight partners and real use cases — lending a solidity that goes beyond the hype and firmly places this technology on the map of innovations that will shape the next decade of science. 🔬
