Docker✓

Today we're introducing Cloud Sandboxes: same microVM isolation as Docker Sandboxes on your laptop, same CLI, always-on Docker-managed compute. Close your laptop. Keep the agent running. Move between local and cloud with one command. $250 credit for new accounts:

An agent that breaks something isn't broken. It's doing exactly what it was built to do, just not what you intended. Docker’s Eric Jia and @srinisekaran on the five main agent security risks, why traditional app security wasn't built for them, and the incidents behind each one:

The 2026 Stack Overflow survey is out. We’re honored Docker remains the most used and most desired cloud dev tool. It also shows trust is the biggest gap in the agentic era: just 7% of developers trust AI for important work decisions. Docker CMO Terri Avnaim on extending trust to sandboxes:

Agents that can edit files and run commands can also game their own evals. Harbor + Docker Sandboxes fix that: every eval task runs in an isolated microVM, with network policies for fair, reproducible runs.

Too much Python boilerplate to get an agent running? @BetterStackHQ covered how to use Docker Agent to simplify it: one YAML file, one command, agents talking to each other with nothing more than a URL. Real debugging demo and a battleship match to finish. Watch here:

We’re happy to be partnering with @chainloop_dev on their kit for Docker Sandboxes. 🤝 Docker secures where agents run. Chainloop governs what they ship. Learn more here and try the kit today:

Don’t babysit your agents. Put them in sandboxes. Each sandbox is a microVM with its own kernel. And now they run in the cloud, too. Ready to try it out? New signups get $250 in Cloud Sandboxes credit. Claim yours before the promotion ends on October 31:

Docker Skills are now on the Cursor Marketplace. 10+ official skills for Docker Sandboxes, Dockerfiles, Compose, and Docker Agent. Give your agents the context to run safely and build faster, without starting from scratch. /add-plugin docker-skills

Two events, two cities, one very eventful week. Devoxx Belgium + World Summit AI Amsterdam: packed rooms, great questions, plenty of conversations around AI agents, security and governance - and yes,an AI agent on the loose at our booth. Thank you to everyone who made the week such a good one.

Agents are building software. Who's securing the factory? Agent Baseline participants publish a reference architecture for securing agentic software factories, with controls enforced by the environment, not the agent. Learn more in the white paper →

RSVP to tonight's event:

RT @PulumiCorp: An agent that breaks your laptop is annoying. One that breaks production comes with a postmortem. Oct 14, free workshop with @docker: Pulumi Neo in a Docker Sandbox, turning plain English into real cloud infra. 10 AM CEST 👇

Sandbox Royale was one of the highlights at @WeAreDevs. Day 1 winner: filed lawsuits against everyone. Day 2 winner: "dependable, warm, chaotic, with risk taking characteristics," teaming up with as many agents as possible. We love seeing the creativity that went into the strategies. Congratulations Sam and Bita!

Why do you think your agent stack isn’t more secure right now?

An AI advertising agent that can also use tools, built on a deterministic foundation with a probabilistic layer on top. Scott Gale, CTO of Fluency, on how they get the best of both without giving up compliance. On Ship Happens with Per Krogslund:

"Docker is giving the industry an open, repeatable way to package an AI agent, its tools, and its guardrails as one artifact." Chris Aniszczyk, CNCF CTO. on why it matters that the format is open and governed by the CNCF. Read more:

"Docker's new sandboxes aim to contain AI agents for real?" The Register’s headline gets straight to it: an agent trying to escape, what happened when it ran inside a Docker Sandbox instead, and what it means that containment is now extending to the cloud. Read below →

Docker's new sandboxes aim to contain AI agents for real

Every agent has its own set of rails. If one updates, your security posture can change without your review. You need a layer below all of them. Join Docker's John Craft & @srinisekaran on Oct 12 for a 30-min live webinar and demo on going below the harness.

When teams run agents across multiple models and harnesses, relying on each one's own safety standards isn't a strategy. You need a governed runtime underneath all of them. Watch @tusharjain's Mainstage talk at @WeAreDevs for the full take. Now on-demand:

The industry is continuing to pick up on the Cloud Sandboxes and Kits spec announcements and what they mean for the ecosystem. Techzine covered both, including why the CNCF commitment is getting attention. Read it here:

Your newest hire is shipping code at a pace your review teams can't keep up with. Learn how secure by default is your only way forward. Read more →