By early 2026, the open source project OpenClaw had become a phenomenon. In January, its GitHub star count crossed 100,000 as developer interest surged. Community dashboards and traffic analytics showed more than 2 million visitors in a single week. By March, OpenClaw topped 250,000 stars, overtaking React to become the most-starred software project on GitHub in just 60 days. Created by Peter Steinberger, OpenClaw is a self-hosted, persistent Artificial Intelligence assistant designed to run locally or on private servers, giving users a way to deploy models without relying on cloud infrastructure or external application programming interfaces.
OpenClaw represents a shift from prompt-based tools to long-running autonomous agents that operate continuously in the background. These agents follow a heartbeat model, regularly checking tasks, deciding whether action is needed and surfacing only decisions that require human input. The approach has also drawn scrutiny around sensitive data handling, authentication, model updates, unpatched local deployments and risks from malicious community contributions. NVIDIA says it is collaborating with Steinberger and the OpenClaw community to improve model isolation, local data access controls and verification of community code contributions, while preserving the project’s independent governance. It has also introduced NemoClaw, a reference implementation that installs OpenClaw, the NVIDIA OpenShell secure runtime and NVIDIA Nemotron open models with hardened defaults for networking, data access and security.
NVIDIA frames autonomous agents as the next wave after predictive, generative and reasoning systems, with each phase increasing inference demand. Generative Artificial Intelligence increased token usage over predictive Artificial Intelligence. Reasoning Artificial Intelligence increased it another 100x. Autonomous agents, which run continuously and act across long time horizons, drive inference demand up by another 1,000x over reasoning Artificial Intelligence. The payoff, according to the company, is the ability to speed productivity by orders of magnitude through overnight research, large-scale design iteration and continuous monitoring that filters out only the anomalies that need human judgment.
The strongest use cases center on always-on workflows, high-iteration tasks and systems that need to move from recommendations to direct action. Examples include monitoring trading systems and regulatory feeds in financial services, updating internal databases from new scientific literature in drug discovery, testing thousands of parameter combinations in engineering and manufacturing, and remediating infrastructure incidents in IT operations. At ServiceNow, Artificial Intelligence specialists leveraging Apriel and NVIDIA Nemotron models can resolve 90% of tickets autonomously.
Responsible deployment depends on auditability, runtime security and local compute. NVIDIA says organizations need visibility into agent reasoning, the ability to inspect and audit actions, and clear intervention points when systems go wrong. NemoClaw is positioned as an open, auditable framework built on OpenClaw’s MIT licensed codebase, with OpenShell providing sandboxed permission boundaries and NVIDIA DGX Spark and NVIDIA DGX Station aimed at sustained local inference with data remaining inside the organization’s environment.
