NVIDIA is releasing the Agent Toolkit, an open foundation for building specialized AI agents that enterprises can custom

The first phase of enterprise AI adoption focused on access, with companies experimenting with frontier and open models through pilots and exploratory projects. Now the industry is moving to a second phase defined by specialized agents — systems of models that can reason, use tools and take action for complex workflows. These agents are already helping life sciences researchers accelerate medicine discovery, security teams investigate vulnerabilities with better context and operations teams coordinate supply chains.

To build effective specialized agents, enterprises need a foundation they can adapt and own, built on customizable models, tools that integrate with existing systems and infrastructure that enables safe operation at scale. NVIDIA Agent Toolkit provides this foundation, comprising models, tools, skills and a secure runtime. It offers an open, modular approach that enterprises and developers can customize, specialize, control and trust. The toolkit works with third-party agent orchestration frameworks including Hermes Agents and OpenClaw.

In life sciences, agents leverage domain-specific models for protein design, virtual screening, genomics analysis and biomarker discovery. The NVIDIA BioNeMo Toolkit enables work that previously took months to complete in days. In healthcare, agents support clinical documentation, clinical decision support and care coordination, while physical robots trained in hospital digital twins can scale surgical assistance and hospital automation. Across software, cybersecurity, industrial operations and customer workflows, agents connect to existing tools and data systems to help teams move faster through complex processes.

Real-world implementations are already underway. Cadence and Synopsys are building autonomous agents for chip design and engineering workflows. CrowdStrike is running specialized security agents that triage alerts with 98.5% accuracy. Palantir, SAP, ServiceNow, Siemens and Dassault Systèmes are embedding agent capabilities into their enterprise platforms where critical business decisions are made. The common pattern across all these examples is that agents become most valuable when they can combine models, tools, skills, runtime and infrastructure in ways companies can tailor to their specific workflows.