Crabbox Quickstart

Crabbox Quickstart

Crabbox runs your repository's commands on a disposable box — a container on your own machine, or a remote machine you lease. It syncs your working tree, runs one command, streams the output back, and exits with that command's code.

The loop

textCopy code
lease -> sync -> run -> read output -> stop

• Lease a box from a provider. --provider local-container uses Docker or Podman on your own machine: no account, no login, no spend.
• Sync your current checkout onto it. Crabbox builds the file list from Git, so the directory must be a repository.
• Run one command there. Its exit code becomes Crabbox's exit code.
• Read the streamed output, and optionally timing and test-result files.
• Stop the box. A one-shot run does this for you; a warm box does not.

Install and check

shCopy code
crabbox --version                  # already installed? skip the install
brew install openclaw/tap/crabbox  # or github.com/openclaw/crabbox/releases
crabbox doctor --provider local-container

doctor with a provider is the readiness check that matters: it names the container runtime, the leases you already hold, and the image= your commands will run inside.

First run, no account needed

With Docker or Podman running, this works in any Git repository with no config file, no login, and no cloud spend:

shCopy code
git init                                  # only if not already a repository
crabbox run --provider local-container -- uname -a

Crabbox provisioned a container, synced the dirty checkout, ran the command there, streamed its output, propagated its exit code, and deleted the lease. Budget 30-45 seconds once the base image is local, nearly all of it container startup; the very first run adds a one-time image pull.

The box is bare

Read this before swapping uname -a for real work. The default image is plain Ubuntu with git, curl, tar, python3, rsync, and passwordless sudo. No node, npm, make, gcc, go, cargo, or java. A command needing a runtime fails before your code does — usually exit 127 and make: not found, or for npm a preflight that stops the run first. Install what you need once, on a warm box:

shCopy code
crabbox warmup --provider local-container          # prints the <slug>
crabbox run --provider local-container --id <slug> -- \
  sudo apt-get install -y make

Commands after -- are literal argv. For &&, pipes, or redirects, add --shell before -- and pass the shell source as one quoted argument. Making the setup permanent is a .crabbox.yaml job — that is the crabbox skill, not this page.

Warm a box and reuse it

One-shot runs pay for container startup every time, and throw away whatever you installed. Keep one box and send several commands to it instead:

shCopy code
crabbox warmup --provider local-container
crabbox run --provider local-container --id <slug> -- ./run-tests.sh
crabbox status --provider local-container --id <slug>

warmup prints both a cbx_... lease id and a friendly slug; either works as --id. A cold run measured 33-45 seconds, the same run warm about 4 seconds. Above and below, ./run-tests.sh stands for your own test command.

What actually gets synced

Your dirty working tree filtered by Git, not a committed ref, so untracked-but-not-ignored files do get uploaded — minus a built-in exclude list: node_modules, dist, target, .venv, __pycache__, and other dependency and build output. That never travels; rebuild it in the box. crabbox sync-plan prints the file count, total bytes, and the largest files and directories without starting a container. Exclude anything else surprising in .crabboxignore.

Errors you will meet first

Exit 2 is a missing --provider, 6 a directory that is not a Git repository, 7 a runtime Crabbox cannot reach, 4 an --id naming no live lease. Your own command's code passes through verbatim too, so the number alone never says which failed — read the message, then rerun the doctor command above.

When your command itself fails, Crabbox exits with its code, prints a failure digest with next: commands, and drops a bundle in .crabbox/captures/. Sync excludes it; add .crabbox/ to .gitignore.

Environment and secrets

Nothing from your shell crosses into the box automatically; forwarding is an allowlist by name, set with --allow-env or env.allow in the repo config. Never put a token on a command line; the crabbox skill covers the rest.

Evidence from a run

Evidence flags attach to any run and cost no extra time:

shCopy code
crabbox run --provider local-container --results-auto -- ./run-tests.sh

If your command writes a JUnit XML report, --results-auto finds it without being told the path and summarizes it in one line:

textCopy code
test results files=1 tests=3 failures=1 errors=0 skipped=1

If nothing writes a JUnit file the flag is a silent no-op.

Stop what you started

Boxes from warmup outlive the command; so do runs given --keep. Stop them:

shCopy code
crabbox stop --provider local-container <slug>
crabbox list --provider local-container

An empty list means no leases remain on that provider. Local containers kept by warmup or --keep require explicit stop; they do not expire on their own.

When you outgrow this page

Stop here and load the full crabbox skill as soon as the task involves any of:

• a repository that already has crabbox.yaml or .crabbox.yaml, including one crabbox init --detect has just written
• any provider other than local-container, or a broker login
• a toolchain the base image lacks that you want present on every run
• named jobs, pools, prewarming, a fresh PR checkout, or Windows targets
• environment or secret forwarding, artifacts, or desktop and UI proof
• a failure that crabbox doctor output does not explain