Wedoany.com Report on Mar 18th, At the GTC conference in Santa Clara, NVIDIA launched the AI Agent Toolkit, an open-source software platform designed to help businesses deploy AI agents capable of autonomous thinking, planning, and action. This move signals NVIDIA's expansion from hardware into enterprise software and has attracted numerous industry partners, including Adobe, Salesforce, SAP, ServiceNow, Siemens, and Atlassian, to sign integration agreements.

The core components of this AI Agent Toolkit include the open-model Nemotron series, the open-agent blueprint NVIDIA AI-Q, open skills like cuOpt, and the newly introduced NVIDIA OpenShell runtime environment. OpenShell provides AI agents with system access and task execution capabilities while establishing guardrails for security, networking, and data privacy.

Addressing the cost concerns of enterprise AI deployment, NVIDIA's AI-Q blueprint employs a hybrid architecture that combines cutting-edge models with lightweight Nemotron models. It is claimed to reduce per-query costs by over 50% while maintaining accuracy. Agents powered by AI-Q have performed exceptionally well on the DeepResearch Bench accuracy leaderboard.

Partners have already begun applying this AI Agent Toolkit to diverse scenarios: Salesforce is building a reference architecture to enable employees to use Agentforce agents via Slack; SAP is utilizing NeMo to allow customers to design custom agents; Adobe is running creative and marketing agents; Siemens has introduced a Nemotron-powered Fuse EDA AI agent for chip design workflows.

In the security and healthcare sectors, CrowdStrike and Cisco provide security integrations, and IQVIA has already deployed over 150 agents in pharmaceutical companies, with plans to integrate Nemotron. The toolkit is available via build.nvidia.com, supporting on-premises deployment as well as cloud platforms like AWS, Google Cloud, and Microsoft Azure, driving the widespread adoption of enterprise AI agents.