Skip to content
kaovilaiPublic

About

podman and docker on macOS, no machine, no remote — both CLIs on one socket over Apple containerization (microVM per container)

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

podm

podman's promise, true on Mac — no machine, no remote, no resident daemon.

No manager daemon, for improved security and lower resource utilization at idle. — podman's own pledge, which its macOS story breaks with a resident VM. podm keeps it: microVM per container, zero daemon RAM at idle.


podm lets the stock podman and docker CLIs manage local containers on a Mac with zero setup and zero VM-machine concepts. It exposes both the Docker Engine API and the podman libpod API on a single unix socket, backed by Apple's containerization framework — the same technology as Apple's container CLI — so every container runs in its own lightweight microVM.

podman ps                     # just works — no podman machine, no CONTAINER_HOST
docker ps                     # just works — no Docker Desktop, no context juggling
podman run --rm alpine echo hello
docker build -t myapp .
podman build --platform linux/amd64,linux/ppc64le --manifest myapp .

Why podm?

On macOS, podman normally requires podman machine — a resident Linux VM holding pinned RAM and a guest disk you have to resize — and docker requires Docker Desktop, a resident daemon plus VM that idles in the gigabytes. podm replaces both with:

  • microVM per container (the Apple container model) — no shared VM to keep alive, no guest disk to manage; each VM dies with its container and memory is overcommitted by macOS like any other process, so idle containers page out instead of pinning RAM.
  • an on-demand daemon — podmd runs under a launchd LaunchAgent and exits when it has zero connections and zero running containers. Idle RAM for the whole stack is 0 MB for podm itself and ≈25 MB for Apple's helper services (measured: container-apiserver 9.8 MB + container-core-images 10.0 MB + container-network-vmnet 4.7 MB, 0% CPU — versus a multi-GB pinned VM).
  • stock CLIs, maximal compatibility — podm's policy is to support all flags and commands of both podman and docker by putting the real CLIs in front of a socket that speaks both API dialects. podm itself is a thin client shim that dispatches to the real CLIs with the right environment. You never outgrow it: every new podman/docker flag works on day one, because it is the actual CLI.

See docs/no-machine.md for the full philosophy, the RAM math, and the hard limits we accept.

Install

Homebrew

The formula is ready (Formula/podm.rb) and the release pipeline publishes to the kaovilai/homebrew-tap tap — it becomes installable with the first tagged release:

brew install kaovilai/tap/podm

Until then, use the source flow below (or a build artifact from Actions).

Requirements

  • macOS 26 (Tahoe) or newer, Apple Silicon (arm64)
  • Apple container 1.6.1+ — brew install container
  • The real CLIs you want to use: podman (brew) and/or docker (brew, client only)

From source

git clone https://github.com/kaovilai/podm.git
cd podm
make                # builds podmd (daemon) + podm (client), debug

A macOS .pkg is also available via make installer (system install to /opt/podm, needs admin; see make installer-help for signed/notarized variants).

The source + podm install flow below is entirely user-space — no sudo needed.

Set up the client (podm install)

One-time bootstrap — installs the podm client, symlinks the podman and docker personas, and records the real CLI paths:

.build/debug/podm install

Add --launchd-template assets/launchd/io.github.podm.daemon.plist.in to also install the daemon's LaunchAgent (on-demand lifecycle: idle-exit + kickstart wake). Without it, start the daemon yourself when needed: .build/debug/podmd & (or podm serve).

Then put podm's bin directory first on your PATH (in ~/.zshrc):

export PATH="$HOME/.podm/bin:$PATH"

podm install places podman → podm and docker → podm symlinks in ~/.podm/bin. The shim dispatches on argv[0], locates the real CLIs (recorded in ~/.podm/client-paths.json), and execs them with CONTAINER_HOST / DOCKER_HOST pointing at podm's socket (unix://$HOME/.podm/container.sock). Because it's the actual podman and docker binaries doing the talking, every flag and subcommand they support works.

Zsh completions

podm install also installs a zsh completion for the podm command: podm-specific subcommands (install, uninstall, status, doctor, serve) are completed directly, and everything else delegates to your installed _podman completion — the mirrored surface is podman's own, so podm reuses it rather than duplicating it. When ~/.zsh/completions/ exists the _podm file lands there (the common fpath convention); otherwise it lands in ~/.podm/completions/zsh/ and podm install prints the fpath line to add before compinit.

podman system connection (secondary or primary remote)

podm install registers a podm podman connection pointing at the daemon socket, so the real podman CLI can select podm explicitly — alongside the local machine, podman machine, or SSH remotes:

podman --connection podm ps          # dial podm explicitly
podman system connection default podm   # or make it the primary remote

Registration never changes the default connection; primacy is your call. podm uninstall removes the connection, podm status reports whether it is registered.

Optional: /var/run/docker.sock compatibility

Tools that hardcode /var/run/docker.sock work with one symlink (one-time sudo):

sudo ln -s "$HOME/.podm/container.sock" /var/run/docker.sock

Optional: docker context

podmd registers a podm docker context automatically. Instead of exporting DOCKER_HOST, you can make it your default once:

docker context use podm

Usage

Personas

podm is a multi-call binary: what it does depends on the name it was invoked as.

Invoked as Behavior
podm podm's own subcommands (below)
podman execs the real podman CLI against podm's socket
docker execs the real docker CLI against podm's socket

podm subcommands

Command Purpose
podm install One-time setup: symlinks, LaunchAgent, client-path discovery, zsh completion, podman connection
podm uninstall Remove symlinks, LaunchAgent, and podm-managed state
podm status Daemon state, socket, LaunchAgent, running containers
podm doctor Diagnose the install: CLIs, podmd binary, LaunchAgent (rendered plist + loaded job), socket, daemon ping
podm version Version and build info
podm serve Run the API daemon in the foreground (what launchd uses)

Environment variables

Variable Default Purpose
PODM_IDLE_EXIT_SECS launchd default Seconds of zero connections and zero running containers before podmd exits (the launchd LaunchAgent restarts it on demand)
PODM_DNS_PORT 2054 Host-side port of podm's embedded DNS resolver
PODM_DNS_UPSTREAM 1.1.1.1:53 Upstream resolver for names the embedded DNS server cannot answer locally: IPv4 literal with optional :port (default 1.1.1.1:53); --dns-upstream flag wins, invalid env warns and keeps the default
PODM_CLEAN_VOLUMES enabled Set false to opt out of removing /lost+found from named-volume EXT4 images (it breaks Postgres initdb; per-volume opt-out label: podm.clean-volumes=false)
PODM_BUILDER_CPUS / PODM_BUILDER_MEMORY 2 / 2048MB vCPUs and memory for the Rosetta (native + amd64-via-Rosetta) builder VM. Memory accepts bytes or k/m/g suffixes. Read at daemon start; takes effect when the builder is next created or recreated
PODM_QEMU_BUILDER_CPUS / PODM_QEMU_BUILDER_MEMORY 2 / 2048MB Same, for the QEMU (emulated-arch: s390x, ppc64le, …) builder VM — emulated builds are CPU-hungry, so big multi-arch projects usually want this higher
CONTAINER_HOST / DOCKER_HOST set by shim Point the real CLIs at unix://$HOME/.podm/container.sock — normally you never set these yourself

Volume sync modes

Named volumes are EXT4 images mounted over virtiofs. Write-durability vs speed is tunable per volume (or daemon-wide):

docker volume create -o sync=fsync mydata   # honor guest fsyncs (durable)
Mode Meaning
nosync default; fastest for write-heavy workloads (~1.5×)
fsync honors guest fsync for durability
full fully synchronous writes

Daemon-wide default: podmd --volume-sync <mode>.

Multi-arch builds

Build for the whole matrix — including architectures your Mac cannot natively or via-Rosetta run — with one command:

podman build --platform linux/amd64,linux/arm64,linux/ppc64le,linux/s390x \
  --manifest myapp .
podman manifest push myapp registry.example.com/myapp:latest

How it works:

  • linux/arm64 — native on Apple Silicon.
  • linux/amd64 — Rosetta 2 translates the x86_64 userland inside the arm64 kernel; build and run both work (uname -m reports x86_64).
  • linux/ppc64le, s390x, riscv64, … — no Rosetta exists for these. Upstream apple/container cannot execute their RUN steps (apple/container#1496); podm solves it by provisioning a second BuildKit builder VM with QEMU binfmt registered, classifying each requested platform as Rosetta-eligible vs QEMU-only, splitting mixed --platform builds across both builders concurrently, and merging the halves into a single image index.

Details and examples: docs/multi-arch.md.

What stays resident after a build

Builder VMs are not torn down when the build finishes — they keep running so the next build reuses the warm BuildKit cache instead of paying a cold start (image fetch + VM boot + BuildKit init, minutes on first run). That means:

  • Each running builder holds its configured size in RAM — 2 vCPU / 2048 MB per flavor by default (PODM_BUILDER_* / PODM_QEMU_BUILDER_* above), more if you sized it up. How many builders are resident depends on what your builds used: an arm64/amd64 build leaves the Rosetta builder, an emulated-arch build leaves the QEMU builder, a mixed-platform build leaves both — and builders started by earlier builds stay resident too, regardless of what the latest build used.
  • The daemon itself won't idle-exit while a builder runs: podmd's idle exit (PODM_IDLE_EXIT_SECS, default 90 s under launchd) counts running containers, and builders are containers. That's deliberate — a stopped daemon would just be woken again by the next build anyway.

To reclaim the RAM, stop the builder(s) — they are ordinary containers:

docker stop buildkit        # the Rosetta (native + amd64) builder
docker stop buildkit-qemu   # only exists if you built emulated arches

Stopping is cache-preserving: the builder's disk (with the BuildKit cache) survives a stop, and the next build starts the same VM again — you pay the boot, not the cache rebuild. Only a replacement (recreating the builder to size it up, or deleting the container) resets the cache. Once no builder runs and no other activity is in flight, podmd itself exits per PODM_IDLE_EXIT_SECS and consumes zero RAM until the next client call wakes it.

The first command after an idle exit pays the daemon's cold start (waking the container apiserver and resuming services), so the shim prints podm: waking podmd (it exits when idle)… on stderr once the wait is long enough to notice — wakes that finish in a fraction of a second stay silent. If the LaunchAgent plist exists but the job is not loaded, the shim says how to reload it (launchctl bootstrap gui/$(id -u) "$HOME/Library/LaunchAgents/io.github.podm.daemon.plist") or run a foreground daemon (podm serve), then hands over to the CLI instead of waiting out its own connect retry.

How it works

podman ─┐                                                ┌─ Docker Engine API (/v1.32–v1.51)
docker ─┼─ podm (client shim, argv0) ──────────────────┤                     one socket,
        │   exec real CLI with CONTAINER_HOST/DOCKER_HOST ├─ podman libpod API (/libpod/*)  both dialects
        ▼                                                 ▼
   unix socket  ~/.podm/container.sock
        │
   launchd (LaunchAgent io.github.podm.daemon) — on-demand, idle-exit → 0 MB daemon RAM
        │
   podmd (Swift/Vapor daemon)
        │
   apple/containerization — one lightweight Linux VM per container
   (container-apiserver, BuildKit builder VMs, Rosetta + QEMU for multi-arch)
Piece What it is
podm Client shim: dispatches by argv[0] (podm/podman/docker symlinks), ensures the daemon is awake, execs the real podman/docker CLI with the right environment
podmd Swift daemon (Vapor) serving Docker Engine API v1.51 + libpod API on one unix socket (~/.podm/container.sock)
launchd LaunchAgent managing podmd — started on demand, woken via launchctl kickstart, exits after PODM_IDLE_EXIT_SECS idle (0 connections + 0 running containers)

The one-socket-two-dialects point: docker, compose, docker-py, test tooling and friends speak Docker Engine API on /v1.x/*; the podman CLI speaks podman REST on /libpod/*. Both hit ~/.podm/container.sock and see the same containers, images, volumes and networks. Deep dive: docs/architecture.md.

Limitations

Honest, inherited-from-the-model limits (details in docs/no-machine.md):

Limitation Status
podman pause / docker pause Not supported — a microVM cannot be frozen; the API returns an explicit error
--privileged Mapped to all capabilities inside the guest VM — there is no true host-privileged
Container resources (--memory, --cpus) Fixed at create time; update only handles restart policies
Memory ballooning Partial — guest-freed pages are not always returned to the host
Platform macOS 26+ (Tahoe), Apple Silicon only — hard boundary from Apple's frameworks
Pods / podman kube play Not available — no pod primitive in apple container
Exotic-arch RUN (ppc64le, s390x, …) Build-time only, via the QEMU builder VM; running such containers is not supported
podman commit, podman diff API routes exist but return "not implemented"
Container startup latency ~1s per container (VM boot) vs ~320ms on Docker Desktop

Comparison

podm podman machine Docker Desktop colima socktainer apple container us/mocker
Model microVM per container, on-demand daemon shared Linux VM shared VM + dockerd shared Linux VM (Lima) microVM per container, resident daemon microVM per container microVM per container
Idle RAM ~0 (25 MB Apple helpers) multi-GB pinned VM GB-scale multi-GB pinned VM ~25 MB helpers + daemon ~25 MB helpers ~25 MB helpers
podman CLI ✅ stock ✅ stock (machine UX) ❌ ✅ (machine UX) ❌ ❌ ❌
docker CLI / Engine API ✅ stock via docker socket in VM ✅ ✅ ✅ (partial: 52/69 endpoints) ❌ ❌ (own CLI only)
libpod API ✅ ✅ ❌ ❌ ❌ (upstream) ❌ ❌
Multi-arch (ppc64le/s390x) ✅ QEMU builder manual manual manual ❌ arm64/amd64 (Rosetta) only ❌ arm64/amd64 (Rosetta) only ❌
No machine concept / lifecycle ✅ ❌ ❌ ❌ ✅ ✅ ✅
License Apache-2.0 Apache-2.0 proprietary MIT Apache-2.0 Apache-2.0 AGPL-3.0

No other project does the union: stock docker+podman CLIs + dual-API socket + microVM-per-container + no resident daemon. See RESEARCH.md for the full landscape study.

FAQ

Why the name podm? Type podm and hit TAB: the shell completes to podman — which is exactly the point. podm is the short, nimble form; podm + TAB is the full podman experience (the persona symlink). And the name states the mission: podman's promise, true on Mac.

Is this a fork? The server (podmd) was imported from socktainer (Apache-2.0) with its podman libpod API layer and multi-arch build support, then rebranded and extended. See NOTICE for attribution of socktainer and Apple code; the license stays Apache-2.0.

Podm spun out because socktainer's PRs could not get reviews — this repo needed a place to move at its own pace. Everything here is Apache-2.0: if upstream would like any of it — the Docker Engine API routes, the libpod dialect work, the stores, the conformance/docgen harness, or the extensive test suite — take it. A cherry or a rebase beats a rewrite; the git history is clean and every commit is signed off.

Does kind / testcontainers / docker-py work? Unverified, and in places genuinely hard: anything needing true host-privileged behavior (kind, k3s-in-container) only gets all-caps inside the guest VM, which is not the same thing. Plain API clients (docker-py, compose, Supabase-style tools that read /var/run/docker.sock inside a container — podm vsock-relays that socket) have the best odds. Test before relying on it; report what you find.

Where does my data live? Everything podm-managed is under ~/.podm (socket, lock, client config) and Apple container's own image/content store. No data in /var except the optional /var/run/docker.sock symlink.

Why not just use container system stop when idle? You can — podm only needs Apple's helpers while containers/builds run and auto-starts them on demand. The ~25 MB helper footprint is the floor the platform gives you; the point is nothing above it stays resident.

Route coverage

185 routes registered (156 supported, 29 deliberately not implemented — every gap carries its reason and its tracking context in RouteSupportManifest.swift). Full, searchable version: podm route coverage site.

Status Method Route Dialect Reason
✅ GET /_ping docker
✅ HEAD /_ping docker
✅ POST /auth docker
✅ POST /build docker
✅ POST /build/prune docker
✅ POST /commit docker
✅ GET /configs docker
✅ POST /configs/create docker
✅ DELETE /configs/{id} docker
✅ GET /configs/{id} docker
✅ POST /configs/{id}/update docker
✅ POST /containers/create docker
✅ GET /containers/json docker
✅ POST /containers/prune docker
✅ GET /containers/{id:.*}/archive docker
✅ HEAD /containers/{id:.*}/archive docker
✅ PUT /containers/{id:.*}/archive docker
✅ DELETE /containers/{id} docker
✅ POST /containers/{id}/attach docker
✅ GET /containers/{id}/attach/ws docker
✅ GET /containers/{id}/changes docker
✅ POST /containers/{id}/exec docker
✅ GET /containers/{id}/export docker
✅ GET /containers/{id}/json docker
✅ POST /containers/{id}/kill docker
✅ GET /containers/{id}/logs docker
🚫 POST /containers/{id}/pause docker Blocked upstream: apple/container exposes no pause XPC route; apple/containerization#366 removed the LinuxContainer/LinuxPod pause/resume API (suspected vended-vsock EBADF asserts), while the underlying VZ pause/resume VM primitives remain (apple/containerization#998).
✅ POST /containers/{id}/rename docker
✅ POST /containers/{id}/resize docker
✅ POST /containers/{id}/restart docker
✅ POST /containers/{id}/start docker
✅ GET /containers/{id}/stats docker
✅ POST /containers/{id}/stop docker
✅ GET /containers/{id}/top docker
🚫 POST /containers/{id}/unpause docker Blocked upstream: apple/container exposes no pause XPC route; apple/containerization#366 removed the LinuxContainer/LinuxPod pause/resume API (suspected vended-vsock EBADF asserts), while the underlying VZ pause/resume VM primitives remain (apple/containerization#998).
✅ POST /containers/{id}/update docker
✅ POST /containers/{id}/wait docker
✅ GET /distribution/{name:.*}/json docker
✅ GET /events docker
✅ GET /exec/{id}/json docker
✅ POST /exec/{id}/resize docker
✅ POST /exec/{id}/start docker
✅ POST /images/create docker
✅ GET /images/get docker
✅ GET /images/json docker
✅ POST /images/load docker
✅ POST /images/prune docker
✅ GET /images/search docker
✅ DELETE /images/{name:.*} docker
✅ GET /images/{name:.*}/get docker
✅ GET /images/{name:.*}/history docker
✅ GET /images/{name:.*}/json docker
✅ POST /images/{name:.*}/push docker
✅ POST /images/{name:.*}/tag docker
✅ GET /info docker
✅ GET /networks docker
✅ POST /networks/create docker
✅ POST /networks/prune docker
✅ DELETE /networks/{id} docker
✅ GET /networks/{id} docker
✅ POST /networks/{id}/connect docker
✅ POST /networks/{id}/disconnect docker
🚫 GET /nodes docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 GET /nodes/{id} docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 POST /nodes/{id}/update docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
✅ GET /plugins docker
✅ POST /plugins/create docker
✅ GET /plugins/privileges docker
🚫 POST /plugins/pull docker The docker plugin runtime (privileged plugin container + /run/docker/plugins socket) has no mapping onto Apple's microVM isolation; the route answers not-implemented for CLI compatibility.
✅ GET /plugins/{name} docker
🚫 POST /plugins/{name}/disable docker The docker plugin runtime (privileged plugin container + /run/docker/plugins socket) has no mapping onto Apple's microVM isolation; the route answers not-implemented for CLI compatibility.
🚫 POST /plugins/{name}/enable docker The docker plugin runtime (privileged plugin container + /run/docker/plugins socket) has no mapping onto Apple's microVM isolation; the route answers not-implemented for CLI compatibility.
✅ GET /plugins/{name}/json docker
🚫 POST /plugins/{name}/push docker The docker plugin runtime (privileged plugin container + /run/docker/plugins socket) has no mapping onto Apple's microVM isolation; the route answers not-implemented for CLI compatibility.
🚫 POST /plugins/{name}/set docker The docker plugin runtime (privileged plugin container + /run/docker/plugins socket) has no mapping onto Apple's microVM isolation; the route answers not-implemented for CLI compatibility.
🚫 POST /plugins/{name}/upgrade docker The docker plugin runtime (privileged plugin container + /run/docker/plugins socket) has no mapping onto Apple's microVM isolation; the route answers not-implemented for CLI compatibility.
✅ GET /secrets docker
✅ POST /secrets/create docker
✅ DELETE /secrets/{id} docker
✅ GET /secrets/{id} docker
✅ POST /secrets/{id}/update docker
🚫 GET /services docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 POST /services/create docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 GET /services/{id} docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 GET /services/{id}/logs docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 POST /services/{id}/update docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 POST /session docker dockerd's buildx session channel is not supported; the route answers not-implemented for CLI compatibility — POST is the method every Docker API version since v1.31 defines, and the one buildx's session dialer issues.
🚫 GET /swarm docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 POST /swarm/init docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 POST /swarm/join docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 POST /swarm/leave docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 POST /swarm/unlock docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 GET /swarm/unlockkey docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 POST /swarm/update docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
✅ GET /system/df docker
🚫 GET /tasks docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 GET /tasks/{id} docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
🚫 GET /tasks/{id}/logs docker Swarm object — Apple container has no swarm mode; the route answers moby's documented 503 for CLI compatibility.
✅ GET /version docker
✅ GET /volumes docker
✅ POST /volumes/create docker
✅ POST /volumes/prune docker
✅ DELETE /volumes/{name} docker
✅ GET /volumes/{name} docker
✅ GET /libpod/_ping libpod
✅ HEAD /libpod/_ping libpod
✅ POST /libpod/auth libpod
✅ POST /libpod/build libpod
✅ POST /libpod/commit libpod
✅ POST /libpod/containers/create libpod
✅ GET /libpod/containers/json libpod
✅ POST /libpod/containers/prune libpod
✅ GET /libpod/containers/stats libpod
✅ GET /libpod/containers/{id:.*}/archive libpod
✅ HEAD /libpod/containers/{id:.*}/archive libpod
✅ PUT /libpod/containers/{id:.*}/archive libpod
✅ DELETE /libpod/containers/{id} libpod
✅ POST /libpod/containers/{id}/attach libpod
✅ GET /libpod/containers/{id}/changes libpod
✅ POST /libpod/containers/{id}/exec libpod
✅ GET /libpod/containers/{id}/exists libpod
✅ GET /libpod/containers/{id}/export libpod
✅ GET /libpod/containers/{id}/json libpod
✅ POST /libpod/containers/{id}/kill libpod
✅ GET /libpod/containers/{id}/logs libpod
🚫 POST /libpod/containers/{id}/pause libpod Blocked upstream: apple/container exposes no pause XPC route; apple/containerization#366 removed the LinuxContainer/LinuxPod pause/resume API (suspected vended-vsock EBADF asserts), while the underlying VZ pause/resume VM primitives remain (apple/containerization#998).
✅ POST /libpod/containers/{id}/rename libpod
✅ POST /libpod/containers/{id}/restart libpod
✅ POST /libpod/containers/{id}/start libpod
✅ GET /libpod/containers/{id}/stats libpod
✅ POST /libpod/containers/{id}/stop libpod
✅ GET /libpod/containers/{id}/top libpod
🚫 POST /libpod/containers/{id}/unpause libpod Blocked upstream: apple/container exposes no pause XPC route; apple/containerization#366 removed the LinuxContainer/LinuxPod pause/resume API (suspected vended-vsock EBADF asserts), while the underlying VZ pause/resume VM primitives remain (apple/containerization#998).
✅ POST /libpod/containers/{id}/update libpod
✅ POST /libpod/containers/{id}/wait libpod
✅ GET /libpod/containers/{name}/healthcheck libpod
✅ GET /libpod/events libpod
✅ GET /libpod/exec/{id}/json libpod
✅ POST /libpod/exec/{id}/start libpod
✅ GET /libpod/images/export libpod
✅ POST /libpod/images/import libpod
✅ GET /libpod/images/json libpod
✅ POST /libpod/images/load libpod
✅ POST /libpod/images/prune libpod
✅ POST /libpod/images/pull libpod
✅ DELETE /libpod/images/remove libpod
✅ GET /libpod/images/search libpod
✅ DELETE /libpod/images/{name:.*} libpod
✅ GET /libpod/images/{name:.*}/exists libpod
✅ GET /libpod/images/{name:.*}/get libpod
✅ GET /libpod/images/{name:.*}/history libpod
✅ GET /libpod/images/{name:.*}/json libpod
✅ POST /libpod/images/{name:.*}/push libpod
✅ POST /libpod/images/{name:.*}/tag libpod
✅ POST /libpod/images/{name:.*}/untag libpod
✅ GET /libpod/info libpod
✅ POST /libpod/local/images/load libpod
✅ DELETE /libpod/manifests/{name:.*} libpod
✅ POST /libpod/manifests/{name:.*} libpod
✅ PUT /libpod/manifests/{name:.*} libpod
✅ GET /libpod/manifests/{name:.*}/exists libpod
✅ GET /libpod/manifests/{name:.*}/json libpod
✅ POST /libpod/manifests/{name:.*}/push libpod
✅ POST /libpod/manifests/{name:.*}/registry/{destination:.*} libpod
✅ POST /libpod/networks/create libpod
✅ GET /libpod/networks/json libpod
✅ POST /libpod/networks/prune libpod
✅ DELETE /libpod/networks/{id} libpod
✅ GET /libpod/networks/{id}/json libpod
✅ POST /libpod/networks/{name}/connect libpod
✅ POST /libpod/networks/{name}/disconnect libpod
✅ GET /libpod/networks/{name}/exists libpod
✅ POST /libpod/secrets/create libpod
✅ GET /libpod/secrets/json libpod
✅ DELETE /libpod/secrets/{name} libpod
✅ GET /libpod/secrets/{name}/exists libpod
✅ GET /libpod/secrets/{name}/json libpod
✅ GET /libpod/system/df libpod
✅ GET /libpod/version libpod
✅ POST /libpod/volumes/create libpod
✅ GET /libpod/volumes/json libpod
✅ POST /libpod/volumes/prune libpod
✅ DELETE /libpod/volumes/{name} libpod
✅ GET /libpod/volumes/{name}/exists libpod
✅ GET /libpod/volumes/{name}/json libpod

Project status

Early stage. The daemon's API surface (Docker Engine + libpod, multi-arch builds) is imported and working; the client shim and launchd on-demand lifecycle are the new work landing now. Known gaps: podman healthcheck run, manifest annotate, all-container podman stats, and 17 of 69 Docker Engine endpoints still partial (upstream socktainer #14).

Roadmap (ideas, unordered)

  • podman healthcheck run (exec-based on-demand probe)
  • manifest annotate, all-container /libpod/containers/stats
  • Docker Engine API parity completion (socktainer #14, #90)
  • Exotic-arch run (qemu-user registered inside container VMs)
  • Optional pods modeled as multi-VM + shared vmnet + shared volumes

Contributing

See CONTRIBUTING.md. DCO sign-off required on every commit.

License

Apache-2.0 — same as socktainer. See LICENSE and NOTICE.

About

podman and docker on macOS, no machine, no remote — both CLIs on one socket over Apple containerization (microVM per container)

Resources

Code of conduct

Contributing

Security policy

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages