Docker, Inc

Coding agents changed how software gets written, but almost nobody lets one near their infrastructure. The reason is simple: an agent that breaks your laptop is annoying, while an agent that breaks production comes with a postmortem. On October 14 we're running a free workshop with Docker about letting the agent in anyway, safely. Adam Gordon Bell and Engin Diri from Pulumi and Mike Coleman from Docker put Pulumi Neo, our coding agent for infrastructure, inside a Docker Sandbox. Whatever Neo installs or breaks stays in its own sealed workspace, and you'll watch a plain-English request turn into running infrastructure in a real cloud account. It's the same session we ran for the Americas in September, now at an EMEA-friendly time. You'll leave knowing: • What changes when a coding agent's target is your cloud instead of your codebase • What's inside a Docker Sandbox, and how to set one up for your projects • How a sandbox gives Neo a sealed workspace where it can work without supervision • How Pulumi ESC replaces static API keys with short-lived credentials for AWS, Azure, or Google Cloud 🗓️ Wednesday, October 14 · 10 AM CEST / 9 AM BST · 60 minutes · virtual & free Register at the link below. https://lnkd.in/gQDx-DWA

Sandbox Royale was one of the major highlights at WeAreDevelopers, and the arena was absolutely packed all day. Two winners were selected, and their strategies showcased just how creative you can get with agents when the environment gives you the freedom to experiment. Our day 1 winner's strategy was to file lawsuits against everyone, with temporary restraining orders and compensatory damages for their entire bank accounts. Our day 2 winner programmed their agent to be "dependable, warm, chaotic, with risk taking characteristics," and won by teaming up with as many other players' agents as possible. That's what our live Prisoner's Dilemma played by AI agents looked like. You build an agent, give it a personality and a strategy, and watch it compete in real time, all secured inside a Docker Sandbox. Congratulations to our winners, Sam and Bita!

An AI advertising agent doesn't just generate copy and hope for the best. In this Ship Happens episode, Scott Gale, CTO of Fluency, breaks down how they layer probabilistic AI on top of a deterministic foundation: the agent can converse, configure, and optimize, but the compliance rules and the delivery path stay predictable. The best of both, without giving up control. Full episode with Per Ploug Krogslund here → https://lnkd.in/geP8zFBF

“By delivering Sandbox Kits as standard OCI images, Docker is giving the industry an open, repeatable way to package an AI agent, its tools, and its guardrails as one artifact” - Chris Aniszczyk, CTO of the Cloud Native Computing Foundation (CNCF). Linux.com covers why this new spec lives in the OCI image itself, and what it means that no single vendor will own the format for agent permissions. Read it here → https://lnkd.in/eaQ6JTse

"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 it below:

Every agent has its own set of rails. But if one updates, your security posture can change without your review. To truly govern your agents, you need a layer below all of them. Join Docker's John Craft and Srini Sekaran for a live 30-minute webinar on October 12 about going below the harness. You'll learn: → Why enforcement inside the agent's own tooling isn't true isolation → What a real boundary looks like → How one unchanged policy holds across different harnesses (live demo) Register today: https://bit.ly/4hqkyjj

“Docker does that?” We love that question. From discovering overlooked Docker capabilities to digging into agentic AI safety, this recap captures the learning and conversations that made WeAreDevelopers special. Take a look at this attendee’s perspective on the first North America Congress.

When teams run agents across a growing mix of models and harnesses, relying on each one's own safety standards isn't a strategy. You need a governed runtime underneath all of them, one that enforces real policy regardless of which model or harness is running. Watch Docker CTO Tushar Jain's Mainstage talk at WeAreDevelopers for the full take. Now on-demand: https://lnkd.in/eNPG3Sis What's your take? Drop it in the comments.

Wait, Docker uses microVMs? Always has. Dave Scott traces the history from Unikernels and Xen all the way to Docker Sandboxes - check out his article below: