Stenobird

Episode

Docker AI, what’s new with MCP, Agents, Sandboxes, and more

Podcast
DevOps and Docker Talk: Cloud Native Interviews and Tooling
Published
Apr 7, 2026
Duration seconds
4718
Processing state
processed
Canonical source
https://podcast.bretfisher.com/episodes/docker-ai-what-s-new-with-mcp-agents-sandboxes-and-more
Audio
https://media.transistor.fm/9f9fbfbd/aec78fc6.mp3
JSON
/v1/public/podcasts/devops-and-docker-talk-cloud-native-interviews-and-tooling/episodes/docker-ai-what-s-new-with-mcp-agents-sandboxes-and-more
Markdown
/podcast/devops-and-docker-talk-cloud-native-interviews-and-tooling/docker-ai-what-s-new-with-mcp-agents-sandboxes-and-more.md

Actions

  • POST https://stenobird.com/v1/public/podcasts/devops-and-docker-talk-cloud-native-interviews-and-tooling/episodes/docker-ai-what-s-new-with-mcp-agents-sandboxes-and-more/transcription-requests
    Idempotently request low-priority transcript generation for this episode.
  • GET https://stenobird.com/podcast/devops-and-docker-talk-cloud-native-interviews-and-tooling/docker-ai-what-s-new-with-mcp-agents-sandboxes-and-more.md
    Read the agent-friendly Markdown representation of this episode resource.

Summary

Docker is evolving from a container runtime into a critical security and orchestration layer for AI agents. This episode explores how Docker Sandboxes and Hardened Images provide the necessary isolation and governance for running autonomous LLM workflows.

Topics

  • Docker Sandboxes
  • AI Agents
  • Model Context Protocol
  • DevOps Automation
  • Container Security
  • Hardened Images
  • LLM Orchestration
  • Agentic DevOps

Highlights

  • Main idea: Docker Sandboxes provide a micro-VM environment to execute AI agents with network and filesystem isolation
  • Practical takeaway: Use Docker Hardened Images to reduce the attack surface of your production workloads by minimizing CVE counts
  • Failure mode: Relying solely on LLM-native permission models (like Claude's) is insufficient against sophisticated prompt injection attacks
  • Main idea: The Model Context Protocol (MCP) Toolkit enables dynamic discovery of tools and servers for AI agents within Docker
  • Practical takeaway: Implement Docker Agent with GitHub Actions to automate PR reviews and documentation consistency checks

Chapters

  1. 1:00 Docker's AI Evolution: An overview of Docker's recent release cycle, moving beyond container management into specialized AI tooling and product lines.
  2. 13:00 Docker Hardened Images: A deep dive into the ecosystem of hardened system packages and the availability of free vs. paid catalog options.
  3. 25:05 The Shift to Agentic DevOps: Discussing the rapid adoption of new AI paradigms and the need for tools that manage autonomous command execution.
  4. 37:20 Docker Sandboxes & Security: Exploring the necessity of multiple isolated sandboxes to prevent agents from accessing sensitive host data or cross-pollinating environments.
  5. 43:00 Model Runner & MCP Integration: Technical details on running models via Docker Model Runner and using the MCP Toolkit for tool discovery.
  6. 1:00:20 Automating the SDLC with Cagent: How the Docker Agent and GitHub Actions can automate documentation scans and pull request reviews.
  7. 1:12:20 The Future of AI Governance: Reflecting on the 'early PHP days' of AI and the critical need for infrastructure that segments human access from AI access.