Docker is covering the full range of running AI agents safely - at the Agentic AI Conference: - Panel: Tushar Jain on agent governance - Tutorial: Michael Irwin on securing coding agents - Workshop: Oleg Šelajev on sandboxing workflows Apr 6–9, virtual Register → https://bit.ly/41m6Jt7

The biggest failure in recent supply chain attacks isn't a single vuln, it's implicit trust. Stolen creds poison packages, poisoned packages steal creds, and the loop keeps accelerating. Docker’s CISO breaks down what’s actually happening & what teams can do to reduce risk: https://bit.ly/3NMjF8K

Using AI agents and hitting limits? It’s not the model that’s slowing you down - it’s everything around it. Warp CEO Zach Lloyd and Docker COO Mark Cavage break down the real infrastructure bottleneck & what needs to change to actually scale. Watch more: https://bit.ly/41Imta3

Gemma 4 is now on Docker Hub! Gemma 4 supports a wide range of applications, and Docker Hub - which hosts a curated set of AI models, packaged as OCI artifacts - makes it simple to pull and run. Learn more: https://bit.ly/417cpHw

Docker Model Runner now supports the NVIDIA DGX Station GB300. The hardware is powerful but what makes it practical is the developer experience: pull, run, and iterate on larger models using the same Docker workflow you already know and trust. Learn more here: https://bit.ly/4dX6w7g

Agents in YOLO mode. Safely. Sandboxes are now standalone: run any agent or claw in an isolated microVM with no host access and no cleanup. Full autonomy, real isolation. Just `brew install docker/tap/sbx` and go. Get started: https://bit.ly/41MyYBn

Balancing agent autonomy with control? Catch Tushar Jain’s panel at the Agentic AI Conference covering how to design governance for agent systems, from control mechanisms to auditability and safe scaling. Apr 6 | 9:15 AM PT Register: https://bit.ly/41m6Jt7

New Made for Dev episode with Docker. Oleg Šelajev breaks down securing AI workflows: Hardened Images, VM-based Sandboxes, secret handling, and MCP guardrails. Great for leveling up or getting started with Docker in agent workflows. Watch → https://youtu.be/WeoltpwflsQ

New episode of Ship Happens. Per Krogslund talks with Pipedrive’s Kristjan Elias on how infra teams balance cost, reliability, and security, where AI fits in production, and how to choose between local vs external models. Watch → https://youtu.be/xnEi_FRRKuk

If the upside of AI agents is so obvious, why aren’t more teams scaling them? Zach Lloyd (@warp.dev) and Mark Cavage (Docker) on the real blockers - security, governance, and the infrastructure needed to move beyond local workflows. Watch → https://bit.ly/418h5wN

Sandboxes are now fully standalone. No Docker Desktop required. Just install and go. Now with even faster cold starts, works out of the box with Claude Code, Codex, Copilot, Gemini, and Kiro. Even NanoClaw. Get full agent autonomy. Keep everything that matters safe. More → https://bit.ly/4lZ93jg

Operationalizing MCP requires a simple way to build, run, & share servers with the security and control needed to scale. Join our sessions at @aaif.io #MCPDevSummit to see how that works in practice - covering workflows, control planes, & shipping agent skills. Register: https://bit.ly/4lQDOae

What does it take to build secure autonomous AI agents? Join our workshop at the Agentic AI Conference (Apr 9, 9–11AM PT) to learn sandboxing, isolation, and real patterns for running agents safely in production. Free with code DOCKER100 Register: https://hubs.la/Q048970j0

NanoClaw + Docker just showed up in Forbes’ CIO newsletter. The conversation is shifting - not just what agents can do, but how to run them safely once they’re doing real work. Isolation isn’t just a feature anymore. It’s becoming the starting point. Read more → https://bit.ly/4v2blT8

Skills unlock a new class of repeatable workflows. In this issue of the Docker AI Labs newsletter, see how to use Docker Agent + Model Runner to pull, analyze, and generate a reusable IT news brief - all locally. Read → https://bit.ly/4lVqWzz

“Shift left” looks different when agents are part of the workflow. In this post, Moriah Hara, CISO at Vigilance Cyber Security, breaks down what that looks like in practice - and why containers and secure dependencies are becoming a key trust boundary: https://bit.ly/4rUq5Ra

Instead of one AI doing everything, what if you had a team of agents? One plans. One builds. One tests. This post from Docker Captains Marco Franzon & Esteban Maya shows how Docker Agent + Sandboxes make it possible while keeping everything isolated from your machine. Read → https://bit.ly/4lRrHK2

At RSA? Join Docker + Socket.dev for a happy hour. Socket Firewall is now integrated into Docker Hardened Images, helping filter risky dependencies early. Grab a drink and talk directly with the teams behind it. 🗓 March 25 | 4pm PST Register: https://bit.ly/4dFAvAq

The agent conversation is moving fast. This Forbes piece looks at NanoClaw + Docker from a different angle: not just safer agents, but the start of real infrastructure for agent workloads - isolation, orchestration, and blast-radius control. Read → https://bit.ly/4dG4sAk

Vulnerability management just got more actionable. At the RSA Conference, see Docker + Mend.io’s integration using VEX from Docker Hardened Images - less CVE noise, clearer decisions. Cut through CVE chaos and see the full demo live. 📍 Booth #1443 🗓 Mar 24 | 3PM PT

Trying to safely run coding agents in practice? See how workflows behave under pressure and how to add real guardrails with Docker Sandboxes + MCP, from Docker’s Michael Irwin. Agentic AI Conference (virtual): April 6–10, 2026 | 9AM–3PM PDT Register → https://bit.ly/41m6Jt7

“Put YOLO in a box.” The Register covers NanoClaw + Docker Sandboxes and why agents that mutate everything need isolation that goes deeper than containers. Read → https://bit.ly/4bypj61

A fresh perspective on the NanoClaw + Docker integration. This Techzine piece highlights the core idea: agent-level isolation. Each agent runs in its own container with its own environment and data, containing the blast radius if something goes wrong. Read → https://bit.ly/4bWuePx

Crush has integrated the Docker MCP Catalog directly into its workflow. Pull in MCP tools on demand - databases, cloud services, local models - no manual setup. Crush loads what you need as you go. Get started: https://charm.land/blog/crush-and-docker-mcp/

Another take on the NanoClaw + Docker Sandboxes integration - this time from ZDNET. The piece looks at why the pairing works: a lightweight open-source agent platform plus MicroVM isolation so teams can experiment with agents without risking the host system. More: https://zd.net/4rKEgYP

What does it actually take to scale AI agents beyond a demo? In “The State of AI Agents: Insights From 800+ Builders & Leaders” webinar: learn what teams are actually building, where things break, & how to plan for a future-ready agent infrastructure. Mar 25, 8am PST Register: https://bit.ly/47bVNBS

As NVIDIA GTC wraps up, here’s the buzz: - Tokens are the new commodity. - Agents = huge productivity + real risk, with sandboxes as the “aha” moment to remove that tradeoff - AI is already in use - now it’s about making agents trustworthy and scalable. You can still find us at Booth 3207!

As NanoClaw integrates with Docker Sandboxes, its creator Gavriel Cohen explains the thinking behind it - agents are getting powerful enough to do real work. The job of the infrastructure is to contain them so teams can actually put agents to work.

What does enterprise-ready agent infrastructure actually look like? This write-up by The New Stack looks at how NanoClaw's minimal, auditable runtime combines with Docker Sandboxes’ MicroVM isolation to create a layered security model for autonomous agents. Read → https://bit.ly/4uCuURB

Curious how NanoClaw + Docker Sandboxes actually work? This demo shows how to set up NanoClaw in a Docker Sandbox - connect messaging accounts, configure agent tools and access, and run agents inside an isolated MicroVM. Watch the full demo & learn more: https://bit.ly/4rAXoZi

Live from NVIDIA GTC! The conversations this week have been incredible - lots of builders thinking seriously about agent infrastructure, runtime isolation, and how to move AI workflows into production. If you haven’t stopped by yet, the Docker team is still running demos at booth 3207.

As AI agents move into real workflows, traditional security assumptions break down. Our 3Cs framework explains the shift: - Contain the blast radius. - Curate the agent environment. - Control execution in real time. Relevant reading during NVIDIA GTC week: https://bit.ly/4lzS0nP

Ep1 of Made for Dev: Docker series is now live! If you’re curious how developers are using Docker to run AI agents, containerize projects automatically, and secure agent workflows, this first episode is for you.

AI is shipping more code than ever. What’s protecting the system underneath it? On Ship Happens, Per Krogslund talks with Sergey Katsev (Catchpoint) about dependency risk & why sandboxing + isolation are becoming required infrastructure in the agent era. Tune in → https://youtu.be/_u3hZnRf5-w

Lots of great conversations at the Docker booth today. We’re running live demos all day showing how Docker Sandboxes let you run AI agents safely inside isolated environments. If you're at NVIDIA GTC, come by Booth 3207 to watch the demo and talk with the team!

NVIDIA GTC kicks off today! If you're here at the event, come find the Docker team at Booth #3207. We’re talking AI agents, runtime isolation, and what it takes to run agent workflows safely in production. Stop by, say hi, and see what we’re building.

The NanoClaw + Docker Sandboxes story keeps building. Forbes is now covering the integration and why it matters as AI agents move from prototypes to real deployment. When agents install packages and modify files, isolation becomes critical. Read more → https://bit.ly/4cKpr4L

NVIDIA GTC starts tomorrow. Find Docker at Booth 3207 and at the Docker + Clickhouse + RunReveal meetup - March 16, 7:30 PST. Stop by and bring your hardest AI infrastructure questions! See you in San Jose. Register for the meetup here: https://luma.com/xoq2dz0l

The AI agent conversation is shifting from capability to infrastructure. This VentureBeat piece looks at why enterprises are starting to prioritize isolation, containment, and runtime design - and how NanoClaw + Docker Sandboxes fit into that picture. Read more: https://bit.ly/414cgV0

The NanoClaw story keeps getting more interesting. Six weeks ago it was a weekend project on Hacker News. Now it has 22K+ GitHub stars and an integration with Docker Sandboxes to help teams run agents safely. TechCrunch breaks down the story: https://tcrn.ch/4utnBeZ

Everyone’s experimenting with AI agent claws. Not everyone is thinking about isolation. Docker Sandboxes run agents in isolated environments. NanoClaw is already using this model in production. Deploying powerful agents? This changes the risk equation. Read now: https://bit.ly/4bsqVOV

Meet Pradumna Saraf. Docker Captain. Open source advocate. Community builder. In this spotlight, he shares his journey into containers and cloud-native, and what keeps him excited about the future of development. Read more: https://bit.ly/46WNUjF

Claude Code is powerful. But without the right runtime, it can become risky. In the latest Docker AI Labs newsletter, learn how to run Claude Code locally with Docker, connect it to real tools with MCP, and execute safely in isolated sandboxes. Read → https://bit.ly/4bm3wyq

The announcement that NanoClaw can now be run in Docker Sandboxes just went live, and it's trending on Hacker News! Great conversation going here about running agents safely and using microVM isolation. Check it out → https://bit.ly/3PAfM73

The biggest risk with agents isn’t the model. It’s what the agent can access at runtime. NanoClaw now runs inside Docker Sandboxes, giving every agent MicroVM isolation - no host access, smaller blast radius. Safe execution for real-world AI agents: https://bit.ly/3PAIfcU https://bit.ly/4s3f0xX

The March Docker Navigator is out. This issue covers hardened system packages, running DHI in production, sandboxing AI agents, & faster local inference with Docker Model Runner + vLLM on Apple Silicon. If you care about runtime control and CVE clarity, this is for you: https://bit.ly/4bfbdGE

Planning your NVIDIA GTC schedule? Find Docker at Booth 3207. We’re talking about running agents safely, and with more autonomy. Join our sessions & the after-hours meetup March 16 for deeper conversations. Register now for the meetup: https://luma.com/xoq2dz0l

With NVIDIA GTC coming up, everyone is talking agents & autonomy. If you’re running coding agents, isolation matters. Docker Sandboxes runs them in microVMs so you can experiment without reshaping your machine. Run agents freely, safely: https://bit.ly/4unyQph

AI agents are past the demo stage. The latest State of Agentic AI report shows 95% of teams see agents as a strategic priority - and 60% already have them in production. So what’s slowing scale? Security. What does secure-by-default need to look like next? ➡️ https://bit.ly/4bClY76

The AI era demands an era of trust. On The Changelog, Docker EVP Tushar Jain talks raising the bar on trust - from free Hardened Images to Sandboxes that run coding agents in isolated microVMs with scoped access & runtime-injected credentials. 🎧 Tune in: https://bit.ly/4qvQaFv