All pages
Lesson 5.3 · Reach other machines · 10 min
Docker and containers
Containers, images, compose, and the built-in engine.
Manage containers, images and Compose stacks from the sidebar, open a shell in any container, and let the agent start a database for you. If you have no container runtime installed, you can still run containers with the engine built into TRMNLY on newer Macs.
Time: 10 minutes · Works on: macOS app. The built-in engine needs macOS 26 on Apple silicon.
Open the Docker tab
- Show the right sidebar with
⌘⇧\. - Open the More menu and choose Docker.
The same panel is also inside Conn → Containers. Both views share their state, so use whichever is closer.
The panel has four sub-tabs: Containers, Images, Compose and Ports.
Check your runtime
The strip at the top shows one pill per runtime TRMNLY found, each with a status dot. It recognizes Docker Desktop, OrbStack, Colima and its own built-in engine.
- Look for a green dot on at least one pill.
- If a runtime is installed but not running, its pill shows a Start button. Click it.
- If you have more than one runtime, use the menu on the right to choose which one new pulls, runs and Compose commands use.
If no runtime is installed, the panel says No container runtime found and offers install links. On macOS 26 with Apple silicon it also shows a card titled Run containers without Docker. That card is the built-in engine, covered below.
Pull an image and run it
- Open Images.
- Type an image name into the pull field, for example
postgres:16, and click Pull. - When the image appears in the list, click its run button. Its tooltip reads Run a new container from this image.
- In the sheet, set a container name if you want one.
- Under Ports (host:container, comma-separated), enter a mapping such as
5432:5432. The button beside the field, Suggest a free host port, picks a host port nothing else is using. - Add any environment variables under Environment (KEY=value, one per line). Postgres needs
POSTGRES_PASSWORD=devto start. - Turn on Remove on exit (--rm) if you want the container deleted when it stops.
- Run it.
Work with a running container
Open Containers. Each row shows the runtime, image and status. The row's buttons are:
- Start, Stop and Restart.
- Show logs: expands a live log tail under the row. Only one container's logs are open at a time.
- Open shell in focused terminal: runs
docker exec -itin the focused pane, usingbashand falling back tosh. - Remove: force-removes the container after you confirm. Its volumes are kept.
Try it:
- Click into a terminal pane.
- Click the shell button on your Postgres container.
- Run
psql -U postgresinside it, then\qto leave.
Compose stacks
- Open Compose and click +. Its tooltip reads Add a compose project.
- Pick a folder that contains
compose.yamlordocker-compose.yml. - The card lists the stack's services and ports. Use Up, Down and the restart button to control it.
When you open the panel, TRMNLY also looks for a Compose file in the current project's folder and adds it for you. Removing a card only makes TRMNLY forget it. Your Compose file is not touched.
The built-in engine
On macOS 26 with Apple silicon, TRMNLY can run containers itself, with no Docker, OrbStack or Colima installed. Each container boots as its own lightweight virtual machine, and images come straight from the registry.
- On the Run containers without Docker card, click Set Up…. The first setup downloads a Linux kernel. If you already have Apple's
containertool installed, the button reads Set Up (no download) and reuses its kernel. - Once it is set up, TRMNLY appears as a runtime pill. Pull and run images the same way as above.
Know the limits before you rely on it:
- Containers run inside the app. Quitting TRMNLY stops them. Their files stay on disk, so starting one again picks up where it left off.
- Published ports such as
8080:80are forwarded tolocalhostby TRMNLY. - There is no interactive shell. Ask the agent to run commands inside the container instead. It uses
container_exec. - Compose is not supported on the built-in engine.
Let the agent drive
The agent can do all of this through its docker_* tools. Anything that changes state, such as pulling, running, stopping or removing, asks for your approval first. Try:
- "Which container runtimes are running?"
- "Start a Postgres 16 container on a free port with password
dev." - "Show me the last 200 lines of the api container's logs."
- "Bring the compose stack in this project up."
You can ask for the built-in engine by name: "Run Redis on the trmnly runtime."
Check yourself
- A container you started shows in Containers with a running status.
- The shell button opens a working shell in your focused pane.
- Ports shows your container's published port with a container badge.
Go further
- Docker: the panel in full, and the agent's Docker tools
- Connections: more about the built-in engine
- Ports: where container ports sit beside everything else