HeadsUpAI

NVIDIA Releases OpenShell v0.0.55 with Google Vertex AI Support

NVIDIA released OpenShell v0.0.55, an update to its open-source security runtime that isolates autonomous AI agents in protected sandboxes (protected environments). The release adds an inference provider (the backend that runs an AI model) for Google Vertex AI, allowing agents to call Google's models securely. It also introduces profile-backed policy visibility to clarify active security rules.
Version
v0.0.55
New Provider
Google Vertex AI
Security Visibility
Profile-backed policy
Isolation Type
Kernel-level
Container Support
Improved Podman detection

This update builds on the launch of OpenShell by enabling cross-cloud interoperability. By integrating with Vertex AI, NVIDIA allows agents to access external models while maintaining the strict isolation established in previous DNS exfiltration blocks. This bridges the gap between agent goals and the kernel policies governing them.

You can deploy the update via GitHub to run agents against Vertex AI with more reliable Podman and GPU sandbox detection. The release restores standard GPU process filesystem behavior for consistent hardware acceleration. It is available now for secure agent execution in production.

View the full update on github.com
NVIDIA AI
NVIDIA AI
@NVIDIAAI
X

OpenShell v0.0.55 šŸ§© Google Vertex AI inference provider šŸ”’ Profile-backed policy visibility šŸ³ Better Podman detection in the gateway šŸ–„ļø Restored GPU procfs baseline behavior šŸ”§ CI and docs fixes Run agents against Vertex AI, with improved policy visibility and more reliable Podman and GPU sandbox behavior. https://t.co/TRlhGwfF8D

7retweets60likes
View on X

Still wondering? A few quick answers below.

NVIDIA OpenShell is an open-source security runtime designed to protect autonomous AI agents. It uses kernel-level isolation to create sandboxed environments, ensuring that agents can execute multi-step tasks without gaining unauthorized access to sensitive system data or credentials.

The integration allows AI agents running within the secure OpenShell environment to use Google Vertex AI as an inference provider. This means agents can leverage Google's cloud-based models for reasoning and task execution while remaining isolated from the host system's critical resources.

The update introduces profile-backed policy visibility, which provides a clearer view of the security rules currently enforced within the agent's sandbox. This helps developers and security teams verify that the agent is operating within its intended boundaries and following established safety protocols.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards ā†’

Share this update