docker run to Compose

What it's for

Turn docker run commands into a Compose file, so a container you start by hand can be started, stopped and updated with docker compose up -d, and its settings live in a file you can keep.

For example, when a project’s README says to run a long docker run -d -p … -v … -e … command, paste it here and save the result as compose.yaml.

How to use it

Paste one or more docker run (or docker container run) commands, on separate lines or joined with && or ;. Each becomes a service, named after --name or else the image. Lines can be continued with \.

Ports, volumes and mounts, environment variables and env files, labels, restart policy, networks, user, working directory, hostname, entrypoint, capabilities, devices, DNS, extra hosts, memory and CPU limits, ulimits, logging, health checks, GPUs, -i and -t, --privileged, --init, --read-only and more all map to their Compose keys, and anything after the image becomes the command. Port mappings are quoted, as Docker recommends. A named volume (-v data:/path) is also declared at the top level, and a user-defined network is declared as external, since docker run expects it to exist already. The message line lists options that have no Compose equivalent (such as --rm). Copy a link to this input keeps the commands.

Example

Choose Try an example, which pastes an nginx command with a port, two volumes, a time zone and a restart policy. You get:

services:
  web:
    image: nginx:1.27
    container_name: web
    ports:
      - "8080:80"
    volumes:
      - data:/var/lib/data
      - ./site:/usr/share/nginx/html:ro
    environment:
      TZ: Europe/London
    restart: unless-stopped
volumes:
  data: {}

Good to know

Shell variables ($PWD, ${HOME}) are kept as written; Compose fills in ${…} from the environment too, but $(pwd) it doesn’t, so use . instead. Only docker run is read; docker network create and docker volume create lines are skipped. Rare options without a Compose key are listed rather than guessed.