SAN JOSE, California – Alongside new hardware developments, the 2026 edition of Nvidia’s GTC conference was all about agentic AI, as the company unveiled a slew of related offerings to entice the already agent-hungry among us.

Central to Nvidia's agentic-focused ploy was the introduction of yet another of CEO Jensen Huang’s AI scaling laws. Following his 2025 keynote statement that “the world got scaling laws wrong,” Huang used the 2026 event to introduce so-called "agentic scaling," where AI-to-AI communication drives demand for low-latency, large-context inference at scale.

To satisfy the growing appetite for AI agents, then, Nvidia finally revealed its anticipated OpenClaw equivalent.

Dubbed NemoClaw, the worst-kept secret in AI agent development was teased by industry reports in the days leading up to GTC. Upon its official unveiling, the platform was pitched as a means to democratize always-on agents.

Huang was full of praise for the original OpenClaw, whose creator recently took his trade to OpenAI, saying the system “opened the next frontier of AI to everyone.”

"OpenClaw has open sourced the operating system for agentic computers," Haung said during his keynote. "It's no different than how Windows made it possible for us to create personal computers; now open claw has made it possible for us to create personal agents."

The Nvidia CEO even said every company in the world "needs to have an Open Claw strategy."

Huang wasn’t the only company official full of praise for OpenClaw, with Kari Briski, Nvidia’s VP for generative AI software for enterprise, boldly calling it “likely the single most important software release in history” in a pre-event press conference attended by SDxCentral.

The offering allows users to employ any coding agent and run it locally on their dedicated system. It also employs the new Agent Toolkit (more below) to allow for OpenClaw to be optimized via a single command.

Agent Toolkit and OpenShell

The tech giant also reworked its Agent Toolkit, a suite providing access to open source models and software solutions to encourage enterprises and developers to build AI agents.

Among the additions to the platform was OpenShell, an open source runtime that enforces policy-based security, network, and privacy guardrails for agents, or "claws."

To boost guardrails for Agent Toolkit users, Cisco is lending its AI Defense security solution, while CrowdStrike’s Secure-by-Design AI Blueprint embeds protections from its flagship Falcon platform directly into OpenShell and other Nvidia AI agent architectures.

The toolkit product also contains AI-Q, Nvidia’s reference architecture for building AI agents.

The company claims the offering, a blend of industry-standard foundation models and Nvidia's own Nemotron line of open models, cuts query costs by more than 50%.

LangChain, the AI engineering company behind the popular developer framework of the same name, is working with Nvidia to integrate the Agent Toolkit into its deep agent library to allow enterprises to run AI agents at scale.

The pair have also made it possible to run OpenShell with LangChain locally on dedicated Nvidia GeForce RTX PCs and workstations, as well as its Blackwell-powered desktop supercomputers, DGX Spark and Station.

Nemotron AI model expansions

From software suites and crustacean consoles to open AI models, Nvidia expanded its Nemotron portfolio to cater to agentic needs. Chief of which was Nemotron 3 Omni, which combines audio, vision, and language understanding to allow AI agents to gain insights from user videos and documents.

Nemotron 3 VoiceChat, meanwhile, is focused on handling use cases related to conversations. The model is capable of listening and responding simultaneously thanks to underlying support for automatic speech recognition, large language model (LLM) processing, and text-to-speech capabilities.

“Open source AI has become a global force for innovation,” Briski said. “Nvidia open model families extend intelligence beyond language, enabling developers worldwide to build intelligent agents and power breakthroughs across digital and physical industries.”

Nvidia’s Nemotron Coalition

Rounding out the model-related announcements, GTC saw the unveiling of the Nemotron Coalition, with Nvidia bringing together open model builders and AI labs to “bring together their expertise to build open frontier models collaboratively.”

Taking part alongside the aforementioned LangChain are firms including Mistral AI, Perplexity, and Cursor. They’ll work together along with Nvidia to share research, data, and compute to accelerate open progress on AI models.

The Nemotron Coalition’s first project will see it focus on a base model codeveloped by Mistral and Nvidia, which the company said will “underpin” the upcoming fourth-generation family of its Nemotron line of open models.

“The Nvidia Nemotron Coalition unites world-class AI labs to develop frontier open models that champion transparency, collaboration, and sovereignty – broadening access to intelligence and ensuring the future of AI is shaped with the world and built for the world,” Huang said.