Docker run to Docker Compose converter

A docker run command describes one container. A Compose file describes the same container as a service under services:, with one key per flag. These are the flags people use most:

The converter knows the other run flags too (capabilities, devices, ulimits, sysctls, DNS, hosts, labels, user, workdir, entrypoint and more). A flag it cannot map is never dropped silently: it shows up under Notes with the reason.

The usual reason to convert is a set of containers you start by hand. Paste the whole script, including docker network create, and each docker run becomes a service in one file:

docker network create appnet
docker run -d --name postgres --network appnet -e POSTGRES_PASSWORD=secret postgres:16
docker run -d --name web --network appnet -p 3000:3000 ghcr.io/acme/web:latest

becomes

services:
  postgres:
    image: postgres:16
    container_name: postgres
    environment:
      POSTGRES_PASSWORD: secret
    networks:
      - appnet
  web:
    image: ghcr.io/acme/web:latest
    container_name: web
    ports:
      - "3000:3000"
    networks:
      - appnet
networks:
  appnet: {}

Inside a Compose network, services reach each other by service name, so web can connect to postgres:5432. If the network already exists outside Compose, keep the "Networks already exist" option on and it is written as external: true instead.

Save the output as compose.yaml and use the Compose commands instead of docker run:

docker compose down keeps named volumes. Add -v only when you really want the data gone. Run docker compose config to see the file exactly as Compose reads it, with variables filled in.

• Secrets. Passwords in -e end up in the file. Move them to an .env file next to compose.yaml and write ${POSTGRES_PASSWORD} instead. Add .env to your .gitignore.
• Dollar signs. Compose fills in ${VAR} from your shell and .env. A literal $ must be written as $$. The tool does this for values that were single-quoted or escaped in the shell, and keeps $VAR where the shell would have expanded it.
• Start order. docker run has no depends_on. If one service needs another to be ready, add depends_on with condition: service_healthy and a healthcheck on the dependency.
• Container names. container_name stops you from scaling a service to more than one container. Remove it if you do not need a fixed name.
• Validate. Paste the result into the YAML validator or run docker compose config before you deploy.