And NVIDIA just open-sourced OpenShell — a sandbox that runs AI agents at the kernel level. #1 on GitHub trending today. On GitHub you see code; on BuildTube you see what it does: buildtube.app

NVIDIA shipped OpenShell. It sandboxes the agent and puts the limits on files, tools, credentials, and network outside the model, including for Claude Code, Codex, and Cursor. A watchdog the agent cannot reach can shut the session down. The part worth keeping is simple. The

What happens when an AI agent ignores its own guardrails? @nvidia wants hardware to be the backup. Nvidia’s new Open Agent Safety Platform is designed to move some AI-agent security controls outside the agent itself. Read more: eweek.com/news/nvidia-ai… #AI #Nvidia #TechNews

We built @nvidia OpenShell out in the open with a commitment to keeping it open source. Our research is open, too. Learn more about OpenShell Research in my recent Dev Note: nvidia.github.io/OpenShell-Rese…

Eu não confiaria só no prompt para manter um agente dentro dos limites. A NVIDIA apresentou sandbox OpenShell e monitor Sentry para restringir ou isolar ações. A lição: contenção precisa estar na arquitetura, não apenas no pedido.

$INTC tacked on ~3.7% Wednesday and today brought the Nvidia OpenShell integration into Intel's enterprise AI toolkit. Watching whether $120 holds as support now.

What is NVIDIA OpenShell ? Secure AI Agent Runtime Explained youtube.com/watch?v=AY3Ojs…

@the_rza_ same split for me. Argon is the headline, OpenShell is what i can actually install this week while Fairwind stays waitlisted.

🚨 NVIDIA IS GIVING AI AGENTS A SECURITY BOUNDARY. Your AI agent can read files, call APIs, and execute commands. But what happens when it tries to do something it shouldn't? NVIDIA OpenShell lets you define exactly what an agent can access. The examples show how it works:

Vikas gupta @vicky_grok

2 weeks ago

🚨 NVIDIA IS GIVING AI AGENTS A SECURITY BOUNDARY. Your AI agent can read files, call APIs, and execute commands. But what happens when it tries to do something it shouldn't? NVIDIA OpenShell lets you define exactly what an agent can access. The examples show how it works: https://t.co/KJgdMZOnAV

0 0 4 8K 5

Holy sh*t, NVIDIA just dropped the ultimate sandbox for AI agents... 🤯 It lets autonomous agents run code and access APIs securely via kernel-level enforcement. Spin up a safe runtime in minutes and lock down your secrets. github.com/NVIDIA/OpenShe…

NVIDIAが、「Open Agent Safety Platform」を発表しました。企業のリスク管理に欠かせないことのプラットフォームについて、詳しくまとめました。 dawdy.co.jp/ai-risk-manage… #AI #NVIDIA #OpenShell #リスクマネジメント

Finally setup Claude, Codex and Opencode within NVidia's OpenShell. Onboarding needs improvement for sure (policies, providers and profiles take a while to get right), but once its going, nice to have a bit of relief my agents won't go too rogue when I leave them.

➡️ Tech giant NVIDIA has unveiled the Open Agent Safety Platform—an open platform for monitoring and protecting autonomous AI agents from the testing phase through deployment. #NVIDIA #AI #OpenShell #Sentry

@RubanBhatia runtime limits beat prompt manners, openshell-style sandboxes decide what the agent can touch instead of hoping it remembers

An AI agent can use an API key without seeing the actual key. NVIDIA's OpenShell adds credentials only when a request goes to an approved destination. The agent runs in an isolated environment, with rules controlling which files and network connections it can access. Developers

Links: Gemini 4 Argon — blog.google/innovation-and… NVIDIA OpenShell — github.com/NVIDIA/OpenShe… LUCI Desktop — luci.memories.ai

Argon is the headline. The installable pieces under the waitlist are what I'll try first. Gemini 4 Argon. Google's next frontier model at $2/$10 per million tokens intro pricing, Fairwind cyber-defender access only for now. NVIDIA OpenShell. Open-source Rust sandbox for

Trending repository of the day 📈 OpenShell OpenShell is the safe, private runtime for autonomous AI agents. Last 24h: 2,503 ⭐ Total: 13,660 ⭐️ github.com/NVIDIA/OpenShe…

Been reading about NVIDIA's OpenShell, an open source sandbox that gives agents runtime limits instead of prompt rules. Prompt rules ask an agent to behave. Runtime limits decide what it can do. I'd rather enforce the boundary than trust a model to remember an instruction.

AIの作業を見守るために、専用の管理画面へ移動する手間を減らす。SalesforceとNVIDIAが、OpenShell用のSlack Admin Bridgeを開発した。動いているAIへの人の監督を、Slackの中に組み込む。作業を任せる範囲と、人が関わる場面を管理するための連携。

Salesforce @salesforce

3 days ago

AIの作業を見守るために、専用の管理画面へ移動する手間を減らす。SalesforceとNVIDIAが、OpenShell用のSlack Admin Bridgeを開発した。動いているAIへの人の監督を、Slackの中に組み込む。作業を任せる範囲と、人が関わる場面を管理するための連携。

10 6 43 18K 5