Docker Desktop 4.42: native IPv6, built‑in MCP toolkit, and an OCI model packaging CLI

DockerDevOpsAIContainers

Key update

Docker Desktop 4.42 (June 4, 2025) adds native IPv6 support and full IPv6 port mapping, natively integrates the MCP Toolkit into the UI (with Gordon AI agent integration), and ships a new CLI command (docker model package) to turn models (GGUF etc.) into OCI‑compliant artifacts you can push to registries. The release also re‑enables seccomp by default, tightens various container/kernel defaults, expands Model Runner platform compatibility (including Windows on Qualcomm/ARM GPUs), and improves model discovery/logging in the Desktop UI.

Why it matters

  • Networking parity: Native IPv6 and proper IPv6 port mapping remove brittle workarounds in enterprise and multi‑stack environments. If your infra runs IPv6 or dual‑stack, you can now test and map ports without the previous translations; validate CI pipelines and port‑mapping assumptions after upgrading (some workflows may observe different host↔container addresses).
  • Security posture: Seccomp is re‑enabled by default — good for hardening local dev environments and preventing syscall escapes, but it can break images/benchmarks that relied on unconfined containers. Audit any images or tooling that previously used unsecured profiles and add explicit exemptions where necessary.
  • AI/ML lifecycle: docker model package standardizes packaging models as OCI artifacts. That lets teams reuse existing CI/CD, image scanning, signing, and registry workflows for models — improving provenance, reproducible deployments, and entitlements. If your org delivers models to production, start evaluating packaging models as OCI images and incorporate model scanning/signing in your pipeline.
  • Tooling ergonomics: MCP Toolkit + Gordon integration brings 100+ MCP servers and model discovery into Docker Desktop. That accelerates experimentation (pull/test models quickly from the Desktop UI), but it also increases surface area: validate registry and access policies, and restrict model pull permissions in shared environments.
  • Operational impacts: disk sizing defaults, enhanced logs for Models, and Docker Hub model cataloging reduce friction when onboarding models locally. Also note platform changes (e.g., Mac virtualization options deprecations and new Windows ARM model support) — test developer images on target host OSes to avoid surprises.
  • Immediate actions for teams:
    • Upgrade Docker Desktop on dev machines and CI runners (or test in a canary group).
    • Run smoke tests for IPv6 and dual‑stack scenarios; verify port mappings and DNS behavior.
    • Audit seccomp‑sensitive images and adjust security‑opt only where strictly required.
    • If you ship models, prototype packaging with docker model package and push to a private OCI registry; add scanning/signing into the model CI flow.
    • Review Registry Access Management and sign‑in enforcement if multiple orgs share machines or registries.

Source

Read Next