The everyday work of Docker Desktop. Without Docker Desktop.
A lightweight desktop client for any Docker daemon. Containers, images, volumes, networks, compose and logs in one fast window. Built for Docker in WSL, happy with any daemon you can reach.
Linux AppImage · Windows on ARM · macOS via npm · Node 20+ · Version 1.119.0
Works with any Docker daemonWindows · Linux · macOS

docker ps- Angular
- Electron
- TypeScript
- Node.js
- MIT licensed
- Docker Engine
- Docker in WSL
- Docker over SSH
- Docker Desktop
- Swarm
- Compose
Principles
Talks to your daemon. Doesn’t replace it.
Docker Client MX never installs or manages the engine. It finds your daemon the same way the Docker CLI does and gets to work.
Docker in WSL, with a proper window on it.
Designed around Docker running inside WSL on Windows. Switch Docker contexts straight from the status bar when you need another host.
Everything is one Ctrl+K away.
A command palette for every page and action. Cmd+K on macOS. Your mouse can take the afternoon off.
Everything you open Docker Desktop for.
Nineteen pages, one window. Each one does the job you came for, then gets out of the way. Or see everything it does on one list ↓
Dashboard →
Engine status, CPU, memory and network charts, disk usage, a prune dialog, and what needs attention before it goes wrong.
Containers →
Grouped by compose project, with bulk start, stop, restart and remove. Live logs, a terminal, a file browser and one-click diagnosis on every container.
Images →
Pull with per-layer progress, run a container from an image, see newer versions by patch, minor and major with their release notes, scan for known vulnerabilities, and recreate compose containers.
Volumes & networks →
List, remove and prune. Look inside a volume and change its files, copy it, and see who uses it. Rename a network, and find every port on one page.
Backups →
A volume on a schedule at a fixed time, checked after writing, encrypted with your password and sent to SFTP, S3 or WebDAV. For a database, its own dump. Checked again before it restores, as a whole or only the files you pick, also a whole compose project.
Compose →
Bring whole projects up or down together, as one group, see one as a diagram, check its resilience, and take one to your own server. Turn a docker run command into a compose file, or deploy one from a git repository.
Catalog →
Ready-made compose projects from Arcane, Portainer or a list of your own: fill in a few fields, see what each asks of your engine, and deploy it. Your own projects become templates to share.
Unified logs →
The logs of all running containers merged into one view, with a level filter and filters you save under a name.
Terminals →
A root shell on the Docker host, next to a shell inside any container, and a cheat sheet that puts a Docker command on its prompt.
Claude →
A Claude tab on every container and on the host terminal, with Claude or your own model. It reads logs, events, stats and inspect data and suggests a fix, and acts only if you let it.
AI →
Claude or a model on your own machine. It suggests until you let it act within your rules, helps when a build fails, and lets Claude Code in through an MCP endpoint.
Command palette →
Ctrl+K jumps to any page or container and runs the actions on it. Type an instruction in plain words — the simple ones the app understands by itself.
Connections →
Docker in WSL, a server over SSH, or Docker Desktop. Switch machines from the status bar, and see them all on one page.
WSL →
Docker in WSL without Docker Desktop: move over in one wizard, the .wslconfig with a preview, a disk file that shrinks, ports for your network, DOCKER_HOST for Windows, and WSL after sleep.
Swarm →
Services per stack, nodes, secrets and configs on a manager. Scale, roll out and roll back, deploy stacks from their compose file, and start or join a swarm on your own servers.
Checkup & reports →
One page that checks your setup against thirty-two rules and links to the fix for each, also on a schedule over every machine you follow. Plus a report of what runs on a machine, yours or another, as Markdown or HTML.
History →
The notifications you missed, a day of CPU and memory per container, and one page with everything the engine reported.
Updates →
New images on a moment you pick, per project, also to a newer patch or minor, the database before the app. The app waits to see the container come back and puts the old image back if it does not.
Background mode →
Runs in the system tray, notifies you when a container crashes, becomes unhealthy, hits its memory limit or sets off one of your own rules, also on your phone if you like, and can start at login. In English, Dutch, German, French or Spanish, light or dark.
All 281 things it does, on one list.
Every feature page, boiled down to one line per thing it can do. Each line opens its tile on that page, with the whole story.
281 things on 19 pages.
- Engine status and the current context, always in view
- Why Docker does not start, and the fix, in WSL or over SSH
- Live CPU, memory and network charts
- CPU, memory, network and disk over a day, a week, a month or a year
- Every long task in one list, with its step and duration
- daemon.json as a form or as JSON, put back by itself if Docker stays down
- Disk usage per kind, with the ten largest of each
- One card with crash loops, failing health checks and drift, to clear
- Forecasts for memory, restarts, CPU, network, disk I/O and a full disk
- Recently stopped containers and their exit codes
- Prune images, containers, volumes, networks or build cache
- Grouped by compose project, swarm stack, label or your own groups
- Start, stop, restart, pause or remove many, marked with the space bar
- File browser: upload, download, edit and rename
- Logs since the start, the last hour or around a crash
- Change env, ports, volumes, limits and health check, undone if it fails
- A memory limit proposed from a week of measurements
- Restart when unhealthy, with back-off
- Duplicate a container, with a copy of its volumes
- Commit a container to an image
- Copy as a docker run command or a compose service
- Links to the domain behind Traefik or Caddy
- A fixed address per container under docker.localhost, also https
- IP address per network, in the list
- Disk I/O, CPU throttling and process count
- One-click diagnosis for an issue, secrets left out
- The same diagnosis for images and volumes
- Filter like k9s: several words, !word, /regex/
- Pull with per-layer progress and Docker Hub search
- The platform of an image, and one that runs through emulation
- Private registries through your docker login
- Newer versions by patch, minor and major
- Release notes next to every newer version
- Ignore one version or a whole major
- Move to a newer tag, compose file included, with rollback
- A smaller -alpine or -slim of the same version, sizes from Docker Hub
- A better base image according to Docker Scout, from the menu of an image
- The Docker Hardened Image of the same version, vulnerabilities side by side
- Patch the OS packages of an image with Copacetic
- Vulnerability scan with Trivy, decisions per finding, CSV or SARIF
- SBOM as CycloneDX or SPDX
- Where an image was built, whether it is signed, and the publisher's SBOM
- Recreate the containers still on the old image
- The containers that use an image, by name, one click away
- Run a container, with the usual variables for mysql or postgres
- A Dockerfile, compose.yaml and .dockerignore for a folder, like docker init
- Build from a Dockerfile with live output
- Multi-stage target and build ARGs, remembered per folder
- BuildKit checks on the Dockerfile, and a smaller base image in one click
- Every build of the machine, with the time per step and the cache
- Tag, untag and push
- Every layer as files and as a folder tree, with the contents of a file
- Save an image to a file and load it elsewhere
- Create volumes and networks, with subnet and gateway
- Browse and edit the files in a volume
- Back up and restore a volume
- Copy a volume, owners and permissions kept
- A volume as an image to push, and a new volume from one
- Which containers use a volume, and at which path
- Remove and prune what nothing uses
- The containers on a network by name, with subnet and gateway
- Rename a network, everything moved across
- Every published port on one page, and what listens outside Docker
- Compressed .tar.zst with a sha256, read back after writing
- Daily or weekly at a fixed time
- Schedules that run on their own machine
- Keep the newest 1 to 100, and see what was never backed up
- Skip and retry, or pause the containers for a moment
- Encrypted with age and your own password
- To SFTP, S3 (AWS, B2, Wasabi, MinIO), WebDAV or a registry
- SFTP only to a host key you confirmed
- Native dumps of Postgres, MySQL/MariaDB and MongoDB
- Commands before and after, with docker-volume-backup labels
- Checksum, password and contents checked before a restore
- Restore by merging, replacing, or into a new volume
- Restore single files and folders
- A whole compose project at once, on a schedule, also to a server
- A log of every backup, also by hand, with size, place and reason
- Every backup schedule of every machine on one page
- Set up a new backup from the Backups page, in two steps
- Up and down per project
- Edit compose.yaml, .env and env_files in a VS Code editor
- Open the project folder in VS Code, Cursor or VSCodium
- Earlier versions of a compose file, with a diff and a note
- Turn docker run commands into a compose.yaml
- Write running containers as one compose file
- The project as a diagram, to draw.io
- Drift between running services and their compose file
- Resilience check, with the fix as a diff
- Back up volumes, compose files and .env together
- Pull and recreate only what changed
- Rebuild or recreate a single service
- Compose profiles: which join on start, all of them on stop
- Passwords from Proton Pass or 1Password, never on disk
- Init steps by their service, and jobs to start by hand
- docker compose watch from the menu, for some services or with logs
- Deploy from a catalog of templates
- Deploy from a git repository, reviewed first
- Move a project to another machine with a wizard
- Bind mounts become a volume, a config or a folder there
- Ports taken on the other machine found first
- Copy or move, undone when a step fails
- Update the copy there from here
- Template lists from Arcane, Portainer, Lissy93 or your own
- Arcane registry.json and Portainer templates.json v2 and v3
- Search, and filter by tag or list
- Fill in the .env, with random passwords and secrets
- Fixed values shown, read-only
- Reviewed first: privileged, Docker socket, host folders, old volumes
- A clean environment: no token from your computer reaches it
- include, extends and env_file outside its folder refused
- No deploy if the files changed after the review
- Each template its own compose project, kept on its newest version
- Your own projects as templates, secrets left empty
- Share your templates as a registry.json
- All container logs in one stream, by time
- Only errors and warnings
- Search with words, !word, "a phrase" or a regular expression
- The logs of one compose project
- Switch containers on and off
- Each container in its own column, in your order, or lined up in time
- JSON log lines, formatted
- Filter on a field of a JSON or logfmt line
- Stack traces folded into one item
- stderr lines marked
- Up to twenty named filters, per machine, shared, also from Ctrl+K
- A match shown among the lines around it
- Save a log to a file
- A notification on a log line, also on your phone
- A new kind of error line after an update or restart
- Root shell on the Docker host
- Through a short-lived helper container
- A shell in any running container
- As another user, remembered per container
- Debug shell for distroless or stopped containers
- Claude next to the host terminal
- Closing the tab stops the shell
- Cheat sheet of Docker commands to copy or put on the prompt
- Open the host terminal from the card of the machine you are on
- A Claude tab on every container and the host terminal
- Reads logs, events, stats, inspect data and compose projects
- Passwords and tokens masked before anything is sent
- Proposed actions run only when you click
- Proposed commands wait in the prompt for your Enter
- Help with a failed build
- Opus 5.5, Sonnet 5.5, Fable 5.1 or Haiku 4.5, per conversation
- Or a local model: Ollama, Docker Model Runner, LM Studio
- Your API key, encrypted on your computer
- Your own account, tokens per model, a price of your own and a budget
- Ask from Ctrl+K
- Acting within your rules, and MCP, if you want
- Claude with your own API key
- Any OpenAI-compatible model on your own machine
- Test the speed and tool use of a model
- A remote model address needs https and your confirmation
- Diagnosis for containers, images, volumes and builds
- Explains findings and proposes a compose diff
- Suggest-only by default
- Strict or balanced mode when it may act
- Per action, container or project: right away, ask first or never
- docker-client-mx.ai=never keeps a container off-limits
- No remove, prune, volumes, files or exec for the AI
- A notification for every action it takes
- MCP endpoint for Claude Code and other clients
- Ctrl+K to any page, container, image, volume or project
- Recent picks and what needs attention before you type
- Prefixes: c, i, v, n, p, s and >
- Fuzzy search: parts of words, in any order
- → opens the actions of one item
- Every container action from the keyboard
- Straight to the logs around a crash
- Actions on projects, volumes and networks
- General actions: clean up, check for updates, build
- Plain instructions: “restart mysql and stop php”
- Commands of your own with placeholders, onto the terminal prompt
- Nothing destructive by instruction
- The plan shown before it runs
- Ask Claude from the same box
- Switch machine or theme
- Ctrl+Shift+Space from anywhere in Windows
- Finds Docker like the CLI: DOCKER_HOST, context, default
- Switch contexts from the status bar
- Every machine on one page
- Find a machine by name or by what runs on it
- Follow only the machines you want
- Act on another machine without switching
- Compare two machines, copy an image across
- Move everything to another machine in one wizard, containers too
- Docker Engine in WSL, no Docker Desktop needed
- Servers over SSH, with a connection test
- No open TCP port: docker system dial-stdio over SSH
- Unknown or changed host keys never accepted
- Pick the machine it opens on
- Falls back to the next machine, and says why
- The containers and images of wslc, the container tool in WSL itself
- Copy images between Docker and wslc, without pulling anything
- Run containers, and pull, tag, build and push images in wslc
- The settings of wslc as a form, with a restart of the session
- Move from Docker Desktop to WSL in one wizard
- WSL version, distro, network mode and resources on one page
- .wslconfig editor with a preview and a backup
- Compact the disk file of a distro
- A port for your network, firewall rule included
- DOCKER_HOST for programs on Windows
- Restart WSL when it hangs after sleep
- Wake a distro that WSL stopped
- WSL checks: credsStore, DNS, /mnt/c mounts
- One-click fixes for credsStore and the Docker service
- Mirrored networking and Hyper-V firewall checks
- WSL nodes in a swarm on drain
- WSL options only where WSL runs
- Manager, worker or neither on every machine card
- Services per stack, with the logs of all tasks
- Every task, and why one does not run
- Scale, roll out and roll back services
- Resilience check per stack
- Deploy stacks from compose, changes shown first
- Drain, promote and demote nodes, with quorum warnings
- Create and replace secrets and configs
- Start or join a swarm, token fetched for you
- Join tokens, autolock and heartbeat in one overview
- CPU, memory and network of the whole swarm
- Task containers kept with their stack
- Notices when a node goes down or an update stalls
- Swarm checks: quorum, ingress ports, unused secrets
- “scale web to 3” and “drain node2” from Ctrl+K
- Thirty-two rules in three weights, no score
- --privileged, the Docker socket and writable host folders
- Critical vulnerabilities that already have a fix
- Ports open to your whole network
- No restart policy, health check or memory limit
- Memory limits too tight, too loose or overbooked
- Daemon checks from the CIS benchmark
- Rules for a swarm: managers, drained nodes, ingress
- Rules for WSL, only where it runs
- One click to the dialog that fixes it
- On a schedule, over every machine you follow
- An AI explanation and proposal per finding
- Dismiss a finding, undo it later
- Secrets in the environment hidden until clicked
- Private keys, .env files and passwords in the open
- The last 200 notifications behind a bell
- Muted notifications logged too
- The last three in the tray menu
- A day of CPU and memory per container
- Per container, and over all containers
- A hint for memory that keeps climbing, over hours or days
- When the disk will be full
- All engine events on one page
- Filter events by kind, name, project and period
- Automatic updates per project or container
- Update a single container now or at a fixed time, from its menu
- At a fixed time, 04:00 by default
- A missed night caught up once
- Rollback when a container does not come back
- Stops after the first rollback
- Dependencies first, from depends_on or Watchtower labels
- An edited compose file is left alone
- One summary in the morning
- Update now, rollback included
- Release notes in the morning report
- Same tag, or a newer patch or minor, never a major
- Watchtower pre- and post-update commands
- A cool-down for brand-new releases
- A Trivy scan before an update goes in
- Start, stop and search from the tray
- Starts at sign-in, straight to the tray
- Notifications for crashes, health and memory limits
- Early warnings, at most once a day
- ntfy, Gotify, Discord, Slack, Pushover or a webhook
- Auto-heal for unhealthy containers
- Your own rules on CPU, memory, I/O, logs and events, also elsewhere
- Rules that restart or stop a container
- Weekly cleanup of untagged images and build cache
- Every schedule, backup, update and task of the app on one page
- Updates itself on Windows, Windows on ARM and as a Linux AppImage
- English, Dutch, German, French and Spanish
- Light and dark theme
- Remembers your filters
- Right-click a row for its menu, in every table
- Order the sidebar yourself, and hide the pages you do not use
Optional · bring your own key, or your own model
A container fell over? Ask Claude why.
The Claude tab helps you troubleshoot with the context of your containers and logs at hand. It is entirely optional and uses your own Anthropic API key, or a model on your own machine such as Ollama. It suggests, and acts only if you let it. Leave it off and nothing leaves your machine.
Up and running in a minute.
Pick the way that suits your machine. All three talk to the daemon you already have.
Installer
The Setup.exe for Windows, with a build of its own for ARM processors, or an AppImage for Linux on x86_64: make it executable and start it. All of them keep themselves up to date.
npm
Run it straight away with npx, or install it globally. Needs Node 20 or newer.
$ npx docker-client-mx $ npm i -g docker-client-mx
From source
Clone the repository and run the development build.
$ git clone https://codeberg.org/MartijnSchimmel/docker-client-mx.git $ npm install $ npm run dev