CLI

The binary is available as both ast and as-test. Use npx ast <command> (or add scripts to package.json).

ast --version
ast --help
ast <command> --help

Commands

Command	Purpose
ast test	Build, then run specs. The everyday command.
ast run	Run already-built artifacts (skip the build).
ast build	Compile specs to wasm without running.
ast fuzz	Build and run fuzz targets.
ast init	Scaffold config, runners, and example specs.
ast doctor	Validate config, dependencies, and runtime.
ast clean	Remove build outputs, logs, coverage, and crashes.

test builds and runs in one step; run expects artifacts to already exist (useful after a separate build). See Getting Started to scaffold a project.

Common examples

ast test                       # build + run every spec
ast test math                  # the spec named "math"
ast test --parallel            # spread files across workers
ast test --watch               # re-run on change
ast test --mode node:bindings  # run a specific mode
ast test --fuzz                # specs, then fuzzers
ast test --show-coverage       # print uncovered points
ast fuzz --seed 1337           # reproducible fuzz campaign

Shared flags

Available on most run-like commands (test, run, fuzz, build, doctor, clean as noted):

Flag	Applies to	Effect
--mode <names>	all	Run one or more named modes (comma-separated).
--config <path>	all	Use a specific config file.
--enable <list>	build/run/test	Enable features (e.g. coverage, try-as).
--disable <list>	build/run/test	Disable features.
--jobs <n>	build/run/test/fuzz	Run files through an ordered worker pool of size n.
--build-jobs <n>	build/run/test/fuzz	Cap concurrent builds (defaults to --jobs).
--run-jobs <n>	run/test/fuzz	Cap concurrent runs (defaults to --jobs).
--list	build/run/test/fuzz	Preview resolved files/artifacts without executing.
--list-modes	build/run/test/fuzz	Preview configured and selected mode names.

ast test and ast run

test accepts everything run does, plus the build and fuzz options below.

Flag	Effect
--parallel	Run files through a worker pool sized automatically.
--browser <name|path>	Browser for web modes: chrome, chromium, firefox, webkit, or a path.
--create-snapshots	Record snapshots that don't exist yet.
--overwrite-snapshots	Re-baseline snapshots on mismatch.
--no-snapshot	Skip snapshot assertions for this run.
--show-coverage[=all]	Print uncovered coverage points; =all expands nested gaps.
--suite <names>	Filter to matching suite names or slug paths (alias --suites).
--reporter <name|path>	Use default, tap, or a custom reporter module.
--tap	Shortcut for --reporter tap.
--verbose	Keep expanded per-suite/test lines and live updates.
--clean	Disable in-place TTY updates; print final lines only.

Fuzz options (on ast test)

Flag	Effect
--fuzz	Run fuzz targets after the test pass.
--fuzz-runs <value>	Override iteration count (500, 1.5x, +10%, +100000).
--fuzz-seed <n>	Pin the fuzz base seed.

Watch mode

ast test --watch (or -w) re-runs on source or spec changes and exposes interactive keys:

Key	Action
w	Toggle auto-run / manual mode.
space / enter	Retry failing specs.
a	Re-run all specs.
ctrl+c	Stop.

ast fuzz

Flag	Effect
--runs <value>	Override iteration count (500, 1.5x, +10%, +100000).
--seed <n>	Pin the base seed.
--fuzzer <names>	Filter to matching targets (aliases --fuzzers, --suite, --suites).

See Running Fuzzers.

ast init

Flag	Effect
--target <wasi|bindings|web>	Build target.
--example <minimal|full|none>	Example spec template.
--fuzz-example / --no-fuzz-example	Include or omit the sample fuzzer.
--enable <list> / --disable <list>	Toggle features at scaffold time.
--install	Install dependencies afterward.
--dir <path>	Target directory.
--yes, -y	Non-interactive defaults.
--force	Overwrite managed files.

ast doctor and ast clean

Both take --config and --mode. doctor validates the project (Doctor); clean removes configured outputs and, with no --mode, cleans every configured mode.

Selectors

A positional argument selects which files run; --suite filters which suites run within them by name or slug path.

Selectors resolve against your configured input root(s) — the static prefix of each input glob, e.g. assembly/__tests__ — searched recursively, and fall back to the current directory if nothing matched there.

You type	Resolves to
foo	<root>/**/foo.spec.ts (a spec by name)
rfc/	<root>/**/rfc/**/*.spec.ts (every spec under a folder)
rfc/*.spec.ts	<root>/**/rfc/*.spec.ts (your glob, anchored to the test folder)
*.spec.ts	every spec under the root
./rfc/*.spec.ts	matched from the current directory only (leading ./, /, or ~)
a,b	two selectors, a and b

ast test expectation                             # the spec named "expectation"
ast test rfc/                                     # every spec under an rfc/ folder
ast test 'rfc/*.spec.ts'                          # quote globs so the shell doesn't expand them
ast run expectation --suite "expectations/toBe"   # one suite by slug
ast fuzz --fuzzer "addition"                      # fuzz targets by name

Notes:

• A selector that matches nothing prints a WARN naming where it looked; a ./-prefixed miss suggests the test-folder form (did you mean "rfc/*.spec.ts").
• A bare selector that matches under more than one configured input root prints a WARN and runs all of them.
• Selectors intentionally bypass !-negations in input, so an explicit pick always wins.
• Use --list to preview what a selector resolves to without running anything.

Web modes

Web runs need a browser. Set it per mode (runtime.browser) or with --browser, and add --headless to the runner command for CI:

ast test --mode chromium:headless
ast test --mode firefox --browser /path/to/firefox