Docker Sandboxes: Run AI Agents in YOLO Mode, Safely

Docker Sandboxes (sbx) let you run coding agents — and any other workload — in isolated, safe, managed environments. No more worrying about what an agent in YOLO mode could do to your machine, or what a malicious site might access. Eric Jia (Staff Product Manager) and Kevin Wittek (Staff Software Engineer), both part of the team building Docker Sandboxes, walk you through what they are, how to get started, and the tips and tricks to get the most out of them. In this video, we cover what Sandboxes are, how to get started, and the tips and tricks to get the most out of them: Chapters: 00:00 What are Docker Sandboxes 01:00 Why sandboxes, not containers 02:22 Installing SBX and network policies 04:40 Authenticating your agent (with credential caching) 05:20 Running a containerized workload inside a sandbox 07:00 Statefulness: sandboxes remember what you install 07:50 Controlling what goes in and out (filesystem, network, secrets, MCP) 10:45 Managing secrets with SBX 13:52 The --clone flag for stronger isolation 14:54 Multiplexing agent sessions with git worktrees 17:20 Pulling agent changes back to your host 17:50 Connecting a GUI (Claude Desktop, VS Code, Cursor) over SSH 20:45 Shell sandboxes for anything else you want to run 21:20 Kits and mixins: reusable, shareable sandbox configs Learn more: https://docs.docker.com/ai/sandboxes Get help: https://support.docker.com Install & releases: https://docker.com/products/docker-sandboxes Where to find Docker: Docker: https://www.docker.com/ LinkedIn: https://www.linkedin.com/company/docker Bluesky: https://bsky.app/profile/docker.com X: @docker Instagram: @dockerinc #Docker #DockerSandboxes #AIAgents #AIGovernance #Security