Switching from Docker Desktop to Colima on macOS

I recently switched to Colima as my Docker Desktop alternative on macOS — and I’m not going back. It’s lightweight, starts in seconds, and plugs straight into the existing Docker CLI and docker-compose. If you want a faster local container setup with no new learning curve, Colima is worth trying.

Why I Switched from Docker Desktop to Colima

Docker Desktop worked fine for a long time, but a few things started bothering me:

Resource usage — Docker Desktop was eating 2-4GB of RAM even when idle. On a 16GB MacBook, that’s noticeable.
Slow startup — launching Docker Desktop and waiting for the engine to be “ready” took 20-30 seconds. Colima starts in under 5 seconds.
The UI I never use — I do everything from the terminal. The Docker Desktop dashboard was just sitting there consuming resources.
License concerns — Docker Desktop requires a paid subscription for companies with 250+ employees or $10M+ revenue. Colima is free and open source.

I wanted something that runs containers, works with my existing docker and docker-compose commands, and stays out of the way. That’s exactly what Colima does.

What Is Colima?

Colima is a container runtime for macOS (and Linux). It creates a lightweight Linux VM using Lima, and runs Docker (or containerd) inside it. The key point: it exposes the standard Docker socket, so all your existing Docker commands, Compose files, and workflows work unchanged.

It’s not a Docker replacement — it’s a Docker Desktop replacement. The docker CLI stays the same. Only the engine running behind it changes.

Setting Up Colima on macOS

Here’s the full setup:

1. Install Colima and Docker CLI

brew install colima docker docker-compose

This installs Colima (the VM manager), the Docker CLI (the commands), and the Docker Compose plugin. You don’t need Docker Desktop installed at all — if it’s still installed, you can remove it.

2. Start Colima

colima start --cpu 4 --memory 4 --disk 60

This creates a VM with 4 CPU cores, 4GB RAM, and 60GB disk. Adjust based on your machine. On my M1, this starts in about 3-5 seconds.

If you need to change these later:

colima stop
colima start --cpu 2 --memory 8 --disk 100

3. Verify everything works

colima status
docker info
docker run hello-world

If docker run hello-world prints the success message, you’re good. Every Docker command you already know works from this point.

4. Use Docker like before

docker-compose up -d
docker build -t myapp .
docker ps

Nothing changes in your day-to-day commands. That’s the whole point.

Docker Commands Quick Reference

Colima doesn’t replace Docker commands — it just replaces the engine running behind them:

Task	Command
View all containers	`docker ps -a`
Stop all containers	`docker stop $(docker ps -aq)`
Remove all containers	`docker rm $(docker ps -aq)`
Remove all images	`docker rmi $(docker images -aq)`
Clean everything	`docker system prune -a`
Shell into a container	`docker exec -it <container> sh`

Colima-Specific Commands

These are the only new commands you need to learn:

colima start              # start the VM
colima stop               # stop the VM (containers pause)
colima delete             # delete the VM entirely (careful — you lose containers)
colima status             # check if VM is running
colima list               # list all VMs
colima start --kubernetes # start with K8s enabled

My daily workflow: colima start when I begin work, colima stop when I’m done. That’s it.

Resource Comparison

Here’s the rough difference I noticed on my M1 MacBook Pro with 16GB RAM:

Metric	Docker Desktop	Colima
Idle RAM usage	2-4 GB	~600 MB
Startup time	20-30 seconds	3-5 seconds
Background processes	Multiple (Docker Desktop, Docker Engine, etc.)	Single `colima` process
GUI	Full dashboard app	None (CLI only)
Cost	Free for personal, paid for business	Free and open source

These aren’t scientific benchmarks — just what I observed on my machine during normal development. The RAM difference alone made the switch worth it.

Common Issues and Gotchas

A few things I ran into during the switch:

Colima VM not starting after macOS update: After a macOS system update, Colima occasionally fails to start with a cryptic error. The fix is usually:

colima delete
colima start --cpu 4 --memory 4 --disk 60

You lose running containers, but your images rebuild from Dockerfiles and Compose files anyway. It’s a minor inconvenience.

Volume mounts only cover your home directory: By default, Colima mounts ~ into the VM. If your project lives outside your home directory, containers can’t access the files. I keep everything under ~/projects so this wasn’t an issue, but it’s worth knowing. You can configure additional mounts in ~/.colima/default/colima.yaml.

Docker socket path: Some tools expect the Docker socket at /var/run/docker.sock. Colima creates it at ~/.colima/default/docker.sock. If a tool complains it can’t find Docker, add this to your shell profile:

export DOCKER_HOST="unix://${HOME}/.colima/default/docker.sock"

Port conflicts with AirPlay on macOS: macOS Sonoma uses port 5000 and 7000 for AirPlay Receiver. If you’re mapping containers to these ports, you’ll get a “port already in use” error. Either change the port mapping or disable AirPlay Receiver in System Settings → General → AirDrop & Handoff.

Docker Compose v1 vs v2: If you install docker-compose via Homebrew, you get v2 (invoked as docker compose with a space). If your scripts use docker-compose with a hyphen, either update them or create an alias:

alias docker-compose='docker compose'

Final Thoughts

Switching to Colima took about 5 minutes and I haven’t opened Docker Desktop since. My machine runs cooler, containers start just as fast, and I didn’t change a single Dockerfile or Compose file.

If you’re a CLI-first developer who doesn’t need Docker Desktop’s GUI, give Colima a try. The official README covers advanced configuration like Kubernetes, custom VM profiles, and containerd mode.

If you’re also streamlining your local dev setup, check out how I use Cloudflare Tunnel for stable local webhook URLs — another tool that removed a lot of friction from my workflow.

Thanks for reading — and stay tuned.