Docker, Inc

Apple Silicon just got a serious upgrade for local LLM development. vLLM has become a go-to for high-throughput LLM serving, and now Docker Model Runner supports it on macOS. That means you can run MLX models on an M-series Mac with the same OpenAI-compatible API (plus Anthropic-compatible APIs for tool workflows) using the Docker commands you already know. Pull a model. Run it. Docker picks the right backend. Get started with it on Docker Desktop 4.62+ on Mac. Learn more:

Running local models shouldn’t be complicated. Open WebUI now auto-detects Docker Model Runner for a fully self-hosted setup in minutes. Enable Model Runner, start OpenWebUI, and your models are instantly available in the browser - no extra config. Docker runs the models. Open WebUI gives you the interface. Read the walkthrough:

Congrats to Anil Madhavapeddy and your team on making the cover of CACM! This one’s relevant to every Docker user. Read through the articles to explore the technical foundations behind Docker - and the impressive, real-world systems their work helps power.

Docker Captain spotlight: Meet Kristiyan Velkov - Docker Captain, Front-end Tech Lead, author, hiker, and open-source contributor helping devs ship with confidence. From official Docker guides to his book "Docker for Front-end Developers", Kristiyan shares hard-won lessons to aid other developers learning Docker. Want to containerize front-end apps the right way? Learn from someone who's done it, taught it, and keeps improving it. Read his story -> https://bit.ly/4b0bltD

AI isn’t going away. The real question: what actually matters? Docker’s AI PM Nuno Coração answers these questions (including the question from his daughter about whether 'Claude' is a friend of his). He breaks down signal vs. noise - from AI fundamentals to using Claude Code in his daily workflow, the OpenClaw explosion, and the latest updates to Gordon. This edition of the AI newsletter is part strategy, part hands-on experience: an honest look at how the line between product and engineering is blurring fast - reshaping what it means to build software. Read more:

We're so excited for this! Can't wait to see you all there!

Want to run OpenClaw locally without handing over access to your entire machine? This guide shows how to run OpenClaw inside Docker Sandboxes, isolated in a microVM. You get: - Full local model support - Optional cloud models with secure key injection - Network controls and host isolation - A reproducible, shareable setup All in about two commands. If you're building with AI agents, this is what "secure by default" actually looks like. Read more ⬇️

What if your ChatGPT didn’t just suggest pricing changes, but actually updated Stripe, logged the audit trail, and committed reports to GitHub? In this walkthrough, we build a competitive repricing agent: - Built with the Docker MCP Toolkit - Using GitHub, Stripe, Firecrawl MCPs, Sequential Thinking, Context7 SQLite, and NodeJS Sandbox MCPs - Runs locally in containers you control - No glue code or fragile setups This is agentic AI with real-world actions, not just another chatbot prototype. If you’re exploring MCP or building AI agents that interact with real systems, this is your blueprint. Watch the full demo https://lnkd.in/g2Xk5v3w

What does migrating to free DHI actually look like? Medplum migrated their developer-centric healthcare platform to Docker Hardened Images with just 54 lines of code change - and cut out custom hardening, non-stop CVE triage, and image sprawl in the process. If you're looking to reduce vuln noise or simplify your compliance story, check out how they did it:

Containers solved “how do we move software safely?” Now the question is: ‘how do we let AI agents act safely?” This deep dive by CodeSource connects the early days of Docker to today’s shift toward secure-by-default images, MCP catalogs, and trusted runtimes for agents. From 2am CI failures to AI-driven supply chains - this is the bigger story behind why standardization and trust still matter: https://lnkd.in/gbQNaxK9