NVIDIA's GTC 2026 conference kicks off tomorrow in San Jose, and OpenClaw — the open-source always-on AI agent platform — is at the center of the action.

Build-a-Claw at GTC Park

NVIDIA is hosting a hands-on "Build-a-Claw" experience inside GTC Park from March 16-19. Attendees can set up a personalized AI agent in minutes: name it, define its personality, and connect it to tools like calendars and apps. The agents run locally on NVIDIA DGX Spark or GeForce RTX laptops, with cloud compute also available on-site. NVIDIA describes OpenClaw as "the fastest-growing open source project in history."

OpenClaw on the Big Stage

OpenClaw creator Peter Steinberger is joining NVIDIA's pre-keynote panel on agentic AI, alongside LangChain CEO Harrison Chase and PrimeIntellect CEO Vincent Weisser. The session, part of the GTC Live preshow, streams Monday at 8 a.m. PT — three hours before Jensen Huang's main keynote.

Huang has previously praised OpenClaw as "the most important software release probably ever," and NVIDIA has now published a dedicated OpenClaw Playbook for developers building local-first agents on DGX Spark hardware.

A Platform That's Gone Mainstream

Originally known as Moltbot and then Clawdbot, OpenClaw now operates as an independent foundation with support from OpenAI, which hired Steinberger in February to lead its personal agents work. The project's community has continued to grow independently, and NVIDIA's embrace at GTC marks the clearest signal yet that always-on AI agents have moved from hobbyist experiment to enterprise infrastructure.

Jensen Huang's keynote begins Monday, March 16 at 11 a.m. PT and streams free at nvidia.com/gtc.