What it does
At its core, Minimus AI Integrations is a Dockerfile migration prompt - instructions that teach an AI agent to reach for Minimus images whenever it’s asked to create, migrate, convert, or harden a Dockerfile or Containerfile, swap in a secure or distroless base image, or reduce a container’s CVE count. Rather than guessing at a base image and package list, the agent follows a consistent workflow: discover the relevant Minimus image, select the right tag, inspect runtime requirements, resolve dependencies, compose the Dockerfile, then verify the result and its security posture. This turns Dockerfile hardening from a manual, expert-only task into something any AI agent can do reliably. Because Minimus images follow the dev and production two-variant model, agents are guided to use the fully distroless production variant for the final runtime stage while still relying on the dev variant - which includes a shell and package manager - for the build stage. The result is a multi-stage Dockerfile that builds normally but ships with none of the extra tooling that expands the attack surface.Ways to integrate
Minimus AI Integrations can be adopted at whatever scope fits how you work. The one-off prompt is available to everyone, including Community Edition users; installing the integration per repo or globally requires an Enterprise Edition account.Community Edition
One-off prompt
Paste the migration prompt directly into a chat with your AI agent to migrate a single Dockerfile or scaffold a new containerized app on the spot. No setup or account required.
Enterprise Edition
Per repo AI integration
Add the integration to a single repository so every agent working in that codebase automatically follows Minimus conventions for that project.
Global AI integration
Configure the integration once across your organization so all AI agents, on any repository, default to Minimus images and hardening practices.
Generate skills for minicli
Equip agents to create, configure, and inspect your private Minimus images using natural language prompts.
Built for agents, not just humans
Minimus AI Integrations are designed around how agents actually work: image discovery does not require authentication, so agents can look up Minimus images and tags without any setup friction. The full Minimus documentation is also indexed for LLMs, giving agents a reliable, up-to-date reference to reason over instead of relying on outdated training data about container best practices.Supported agents
The following agents are natively supported:- Claude Code
- Cursor
- OpenAI Codex
- GitHub Copilot (minicli only)
- Windsurf (minicli only)