Docker
See and manage your local containers, images and Compose stacks beside the chat, and let Panda investigate them.
Feature Updated 20 Nov 2026What the Docker app is
Docker runs programs in sealed boxes called containers. The Docker app shows what Docker is running on your computer. It lets Panda read logs and fix a container instead of asking you to paste output.
Turn it on
Docker is off until you ask for it. Open Customize in the sidebar, pick Apps, open Docker (on Explore if it is not on Yours yet) and turn on Enabled under Settings. Docker appears in the Apps folder in the dock, and Panda gets its Docker tools.
When Docker cannot be reached, the Docker app says Docker is not running or Docker is not installed. Anything that provides the local Docker socket works: Docker Desktop, Colima, or a plain daemon. There is nothing else to set.
What it shows
| View | What it holds |
|---|---|
| Containers | Everything running and stopped, with live state, plus start, stop, restart, pause and remove. |
| Images | Local images with their layers, pull, run, scan, remove, and prune. |
| Volumes | Volumes with their driver and mount point, create, remove, and prune. |
| Networks | Networks with how many containers are attached, create, remove, and prune. |
| Compose | Compose stacks discovered on the machine, with up, down, restart, pull, logs, and the resolved config. |
| System | Disk usage broken out by images, containers, volumes and build cache, with per-type reclaim and a full prune. |
Reclaiming space asks before it acts, and asks a second time before it touches volumes. That is the one that deletes data rather than cache.
Inside a container
Open a container and you get its live CPU, memory and network stats, along with five tabs:
- Logs, streamed as they are written.
- Terminal, a real interactive shell inside the container.
- Files, a browser for the container's filesystem, with copy in and out.
- Changes, a diff of everything written since the image was built, which is usually the fastest way to see what a container actually did.
- Inspect, the full engine-level detail.
Terminal and Files need the container to be running, so they step aside for Logs when it stops.
What Panda can do with it
Panda can list and inspect everything, read logs, stats and processes, and see what changed in a container. It can run commands and copy files in it, pull, build, tag and run images, and read Compose logs. It can also create volumes and networks, and remove or prune. Anything that deletes goes through your permission prompts.
To take these tools away from Panda without hiding the tab, set Tools to Off on the Docker page. Turning off Actions under Settings there does the same. With a Limited environment, turn on its Docker Config access preset so Panda can reach ~/.docker.
Start from the panel
A container that stopped with an error or turned unhealthy gets a Why did it exit? or Diagnose button. The log view has Explain. Ask Panda at the top of the list puts the containers' state in the message box. Finish it with a request such as "restart the failing api". Diagnose and Explain send straight away with the container's state and recent logs, so the chat starts with the facts.
While Docker is on, PandaOS watches for containers that stop with an error, run out of memory, keep crashing or fail their health check. It does this even with the panel closed. Each problem gets one row under Problems in the Inbox and a banner at the top of the Docker panel. The banner has Diagnose and Snooze, which mutes that container for 30 minutes. The Inbox row opens the container or dismisses the problem. The watching reads Docker's own events on your computer, so it uses no AI and no tokens until you ask Panda. Desktop pop-ups for these rows stay off until you turn them on in the Inbox settings.
Scan images and Dockerfiles
Panda can scan an image for known vulnerabilities and check a Dockerfile against common mistakes. The scan uses docker scout when it is installed, otherwise a Trivy container. The Dockerfile check runs hadolint in a container.
Both scanners are pinned to a fixed version. The first time one is needed it has to be downloaded, and Panda asks you before it does. After that, scans and checks run without asking.