Docker Sandboxes Crash Course

Isolated dev environments in 1 hour, then a deeper dive into AI agent sandboxing — kits, environment files, the MCP gateway, and multi-agent workflows.

ជំពូកនេះមិនទាន់ត្រូវបានបកប្រែជាភាសាខ្មែរនៅឡើយទេ — សូមអានជាភាសាអង់គ្លេសខាងក្រោម។

A free, self-contained crash course that builds one continuous, hands-on session from start to finish: install Docker Sandboxes, create a disposable sandbox, explore it, run an app inside it, drive Docker from inside it, wire up a small multi-container backend, debug the problems you'll actually hit, and finish with a challenge.

Since there's no live cohort for the deeper material, the course keeps going, free: running a real AI coding agent in a sandbox, packaging setup into reusable kits, declaring a project's whole sandbox in an environment file, giving agents safe tool access through the MCP gateway, and running several agents at once.

Source code and video

The full source and example project referenced throughout this series are available on GitHub: docker-sandbox-crash-course.

  1. 01Docker Sandbox ConceptWhat Docker Sandboxes actually are, what they were built for, and how sandbox isolation differs from container isolation.
  2. 02Getting StartedInstall the sbx CLI, sign in, and create, exit, re-enter, stop, and remove your first sandbox.
  3. 03Exploring the SandboxTour the sandbox's filesystem, user, installed tools, processes, and networking, and prove the isolation is real.
  4. 04Running ApplicationsBuild a small Node.js HTTP server inside the sandbox, publish its port, and reach it from your host.
  5. 05Docker Inside the SandboxUse the sandbox's own nested Docker daemon to pull, run, inspect, and remove a container.
  6. 06Sandbox + Docker Development WorkflowWire up a Node.js backend, PostgreSQL, and Redis as separate containers talking over the sandbox's own Docker network.
  7. 07Common Problems & DebuggingDiagnose a sandbox that won't start, a container that exits immediately, port conflicts, network issues, and permission problems.
  8. 08Final Challenge & RecapRebuild the whole course from scratch in a brand-new sandbox, and recap everything you've learned.
  9. 09Bonus: What Docker Sandboxes Was Actually Built ForRun a real AI coding agent inside a sandbox and see what it can and can't touch, plus when to use shell mode vs. agent mode.
  10. 10Sandbox Kits & Custom TemplatesPackage tools, environment variables, credentials, and agent memory into a reusable kit, then apply it to any sandbox with one flag.
  11. 11Environment Files & SecretsStop retyping CLI flags — declare a project's whole sandbox setup, secrets, and MCP servers in one sbxenv.yaml file.
  12. 12The MCP Gateway for AgentsRegister MCP servers once on the host and let every sandboxed agent reach them, without ever handing an agent your API keys.
  13. 13Multi-Agent WorkflowsRun several agents against the same project safely with clone mode, and see where solo-developer sandboxing ends and organization governance begins.