Secure runtime layer
NVIDIA OpenShell is an open source runtime to build and deploy autonomous, self-evolving agents more safely. It governs how the agent executes, what the agent can see and do, and where inference goes.
Globster uses NVIDIA OpenShell as the secure runtime layer for hosted Luna agents. OpenShell governs execution boundaries, policy enforcement, and inference routing so enterprise teams can deploy autonomous agents with explicit runtime control.
Why this matters
Autonomous agents can write code, spawn sub-agents, use tools, and operate across long-running sessions. As capability grows, application-layer risk grows with it. Enterprises need safety, capability, and autonomy at the same time.
The browser tab model for agents
OpenShell applies a familiar isolation model to agent workloads: sessions are isolated, resources are controlled, and permissions are verified by the runtime before any action takes place.
OpenShell foundation
OpenShell adds security in the environment, rather than relying only on the model or application layer.
Purpose-built for long-running, self-evolving agents. Sandboxes handle skill development and verification, programmable system and network isolation, and execution environments agents can break without touching the host. Policy updates can happen live at sandbox scope as approvals are granted, with an audit trail of allow and deny decisions.
Enforces constraints across filesystem, network, and process layers. Self-evolving agents need granular oversight when they install packages, learn skills at runtime, and spawn scoped sub-agents. The engine evaluates actions at the binary, destination, method, and path level so an agent can use a verified skill without executing an unreviewed binary.
The control point where autonomous agent actions are evaluated before they reach the host. It translates policy rules into runtime enforcement decisions, governing what an agent can read, write, execute, access over the network, or delegate to external tools.
Within Globster
Globster separates transformation, agent composition, and fleet operations. OpenShell provides the execution boundary underneath Luna.
Luna agents can take consequential actions across tools, files, APIs, and workflows. OpenShell provides the runtime boundary that separates agent behavior from host access, so autonomy can increase without handing every agent broad environment permissions.
Globster's control plane defines organization identity, approved models, plugins, budgets, and lifecycle rules. OpenShell enforces runtime policy at the moment of execution: what the agent can see, change, call, and route to inference endpoints.
Security and platform teams need evidence, not assumptions. OpenShell's policy layer produces allow and deny decisions operators can review, aligning with Globster's goal of governed agent fleets rather than unmanaged experimentation.
Enterprise trust
OpenShell is designed to separate agent behavior, policy definition, and policy enforcement. Organizations gain a unified policy layer to define and monitor how autonomous systems operate, regardless of host operating system.
Common questions
Learn more from NVIDIA at docs.nvidia.com/openshell.
No. OpenShell sits underneath agent frameworks and harnesses. Globster's Luna Dynamic Harness remains the composition layer for models, memory, plugins, policies, and workflow logic. OpenShell is the runtime foundation for secure execution on hosted Luna agents.
No. Sandboxing isolates a process for a bounded task. OpenShell is a runtime that governs agent actions across their lifetime, including gateway-brokered credentials, identity boundaries, policy-as-code, policy-enforced egress, inference routing, audit state, and boundaries for multi-agent workflows.
No. OpenShell integrates with the enterprise ecosystem. Identity providers, secret stores, observability systems, security tooling, and governance platforms remain important surrounding systems. OpenShell's role is to provide the agent runtime boundary and enforce policy consistently.
Containers and pods are runtime substrates. OpenShell can use Docker, Podman, Kubernetes, and VM isolation, but adds agent-specific control: gateway coordination, sandbox supervision, policy-enforced egress, credential handling, inference routing, and logs designed around the actions agents take.
OpenShell is a secure runtime for autonomous agents. NemoClaw is a reference blueprint family that packages a harness, model, and runtime together. The runtime layer in NemoClaw blueprints is OpenShell. Globster uses OpenShell directly for the hosted Luna runtime rather than presenting NemoClaw as a separate product surface.
Globster uses NVIDIA OpenShell as the secure runtime layer for hosted Luna agents. OpenShell governs execution boundaries, policy enforcement, and inference routing so enterprise teams can deploy autonomous agents with explicit runtime control.
Next step
Talk with Globster about your runtime requirements, policy model, and the evidence your security team needs before production rollout.