NVIDIA (NASDAQ: NVDA) announced the launch of its Agent Toolkit, an open source software platform for building autonomous enterprise AI agents, according to a company press release.

The toolkit includes NVIDIA OpenShell, an open source runtime that provides policy-based security, network and privacy controls for autonomous agents. The platform also features NVIDIA AI-Q, an agent blueprint that uses a hybrid architecture combining frontier models for orchestration and NVIDIA Nemotron open models for research tasks.

According to the company, the AI-Q hybrid approach can reduce query costs by more than 50% while maintaining accuracy. NVIDIA developed the top-ranking AI agent on the DeepResearch Bench and DeepResearch Bench II leaderboards using the AI-Q Blueprint.

Multiple enterprise software companies are integrating the Agent Toolkit into their platforms. Adobe will use the software for creativity, productivity and marketing agents. Salesforce is incorporating NVIDIA Nemotron models to enable customers to build AI agents using Agentforce. ServiceNow's AI Specialists leverage the toolkit and AI-Q Blueprint with a combination of closed and open models.

Other companies adopting the technology include Atlassian for its Rovo AI system, Box for enterprise file system agents, Cadence for semiconductor design tools, and SAP for AI agents through Joule Studio on its Business Technology Platform.

"Claude Code and OpenClaw have sparked the agent inflection point — extending AI beyond generation and reasoning into action," said Jensen Huang, NVIDIA's founder and CEO.

The Agent Toolkit is available on build.nvidia.com and can run on various NVIDIA hardware platforms including GeForce RTX PCs, RTX workstations, and DGX systems from multiple manufacturers.