NVIDIA OpenShell is an open-source secure-by-design runtime that sandboxes autonomous agents, protecting data and infrastructure during agent execution. It operates as an infrastructure-layer policy enforcement system rather than constraining models, allowing agents to operate autonomously while maintaining security guarantees. OpenShell runs agents inside isolated sandboxes where resource usage is controlled, file access is restricted to authorized paths, network activity is regulated, and credentials are protected through a privacy router.
Policies are defined declaratively in YAML format, making security configuration auditable and version-controllable. OpenShell verifies permissions before any agent action—tool calls, file operations, or network requests—execute, preventing unauthorized data exfiltration or uncontrolled resource consumption. It provides out-of-process policy enforcement and granular permission controls for multi-tenant deployments.
OpenShell integrates with enterprise security partners including Cisco, CrowdStrike, Google Cloud, Microsoft Security, and TrendAI for aligned runtime policy management across the enterprise stack. The runtime supports deployment on local NVIDIA GeForce RTX PCs, RTX workstations, DGX systems, and cloud infrastructure. For knowledge retrieval, agents running in OpenShell can securely access vector databases like Milvus through authenticated connections, ensuring enterprise data remains protected while enabling RAG-powered reasoning.