CLI tool
qawolf/cli avatar
qawolf/cli

QA Wolf CLI: local flows from a per environment cache, a runner SDK, and five standalone binaries

QA Wolf from anywhere — your terminal, your CI, your AI agent.

3,459 stars140 forksTypeScriptApache-2.0

At a glance

What is it?
A command line client that runs a team's test flows on your machine from a per environment cache, installing their runtime dependencies on the way, while flow authoring, AI test generation, cloud execution and team features stay on the vendor's platform. It installs through four package managers or as a precompiled binary per platform, ships an agent skill generated from its own command tree, and offers a TypeScript SDK whose verbs name the runner they address so nothing is billed implicitly.
Who is it for?
QA Wolf CLI suits a team already inside the QA Wolf platform that wants its flows to run on a developer machine or inside a CI job rather than only on hosted infrastructure, and who is willing to have an account and an environment identifier in the loop. It does not suit someone looking for a standalone test framework, because the flows, the authoring and the AI generation all live in the platform, so the CLI is one half of a system rather than a tool on its own.
Can I use it commercially?
Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?
Yes. The repository last received commits 1 day ago.
What is it written in?
Mainly TypeScript, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.

From daily snapshots

Star history

  • qawolf/cli3,459 ★ · +11
Star history
Repository2026-09-112026-11-15
qawolf/cli3.4k3.5k

11 Stars in the last 65 days, measured from the daily GitHub snapshot. Compare with other projects

Editorial analysis

The CLI runs flows locally, and everything else lives on the platform

The division of labour is stated in the second paragraph of the README and it defines what this repository is. The CLI runs and manages flows locally. Flow creation, AI powered test generation, managed cloud execution and team collaboration are part of the full platform, not of this package. So the useful way to think about it is a local execution client for flows that were authored somewhere else, with a public API command for the hosted side as well. Getting started needs an account, and one piece of configuration you cannot invent: the environment identifier, which comes from the platform dashboard under Settings, then Environments. Everything local hangs off that identifier, including the cache directory.

Four package managers, a no-install trial, and five binaries per release

Installation is offered four times over, once per major package manager:

bash
npm install -g @qawolf/cli
# or: pnpm add -g @qawolf/cli
# or: yarn global add @qawolf/cli
# or: bun add -g @qawolf/cli

and there is a fifth route that installs nothing:

bash
npx @qawolf/cli --help

The sixth route sidesteps Node entirely. Each GitHub release carries precompiled binaries, and the README names them: Linux for both 64-bit and ARM, macOS for both, and a Windows executable. The download is a plain fetch and a chmod:

bash
curl -fsSL -o qawolf https://github.com/qawolf/cli/releases/latest/download/qawolf-darwin-arm64
chmod +x qawolf && ./qawolf --help

The runtime note is unusually blunt for a README: Node 20.19 or newer is the minimum, Node 20 reached end of life on 30 April 2026 and no longer receives security updates, and it is supported only for environments still pinned to it, with Node 22 preferred. The binary route sidesteps that decision entirely, since a precompiled executable carries its own runtime, which makes it the practical option for a CI image that has no Node at all.

flows run reads a per environment cache and installs what the flows need

The interesting behaviour is in what a run does before it runs anything. The command takes an environment identifier and works from a directory under the project that holds a per environment copy of your flows on disk. If that copy is not there, the CLI pulls it first, and then it installs the runtime dependencies those flows need. So the first run of an environment is a network operation and a dependency installation, and later runs are local. Two commands bracket that model: one refreshes the cache for an environment explicitly, and the other sets up a local only project so you can author flows without the platform at all. A third command installs the browser runtime once, which the sample flows in the repository's examples directory then expect to be present.

Six command groups, and one of them talks to the public API

The command table is short enough to take in at a glance, and one row explains the scope better than the introduction does. Authentication is its own group. Flows covers running them locally and pulling them. Install covers runtime dependencies, and init covers setting up a project. Doctor exists to diagnose problems, which is the kind of command that tells you the tool expects to be used on machines you do not control. The sixth, run, is the odd one out: it triggers and manages runs on the platform through the public API rather than locally. Every command takes a help flag for its own options, which is where the details the README omits are meant to live. The package manifest matches that shape: one binary, a small set of published files including the SDK build and its type declarations, and a public access setting, so nothing about the distribution is hidden behind a registry login.

Every SDK verb names its runner, so nothing is billed by accident

The runner SDK exists so an agent or a script can drive a runner from TypeScript instead of shelling out, and its contract is unusually careful about billing. The documentation states that every verb names the runner it addresses, so nothing is launched or billed implicitly. Construction takes an API key, a launch call takes a runner identifier and a runner family, and a run call takes an entry point path, an environment, the runner identifier and a selection. The result handling is the part worth copying: a false flag means the request itself failed, covering transport, authentication and local validation, while a runner that refused the work comes back as a success with an outcome and a failure reason taken from a published contract, so a caller can branch over every case without guessing.

Recording keeps a UUID you must reuse, and video URLs expire

Video recording on a browser runner is exposed as another verb, and it has two rules that will bite you if you skip them. First, you generate a recording identifier yourself and pass it with the start action, and the documentation says to keep that identifier when retrying a start or a stop, which implies a retried start with a fresh identifier is a different recording. Second, listing recordings returns one page and takes optional filters for a recording identifier and a page token; the platform URLs it returns are stable, while the video URLs expire. The same verb also accepts a status action and an automatic action that can be switched on or off, and every one of these has an equivalent command line form under the runner group.

A skill generated from the command tree, and three releases in two days

The agent integration is the part of the repository with the most unusual shape. The package ships an agent skill that follows an open specification for such skills and works with several named harnesses, installed through a cross harness installer that detects what is present and asks where to put it, or takes agent flags for a non interactive run. The skill is not written by hand: it is generated from a template plus the command tree by a script, and the test suite is responsible for keeping the two in sync, which is why the build runs the generator before it bundles. The repository carries the supporting machinery for that kind of work, with a changesets folder, custom git hooks wired up by the prepare script, two separate linter and formatter configurations, a configuration file for a dead code checker, a plugins directory and a documentation directory. The release history shows three versions published on the first and second of October 2026.

Editorial conclusion

QA Wolf CLI suits a team already inside the QA Wolf platform that wants its flows to run on a developer machine or inside a CI job rather than only on hosted infrastructure, and who is willing to have an account and an environment identifier in the loop. It does not suit someone looking for a standalone test framework, because the flows, the authoring and the AI generation all live in the platform, so the CLI is one half of a system rather than a tool on its own. Three things to check before you wire it into a pipeline: which Node version your runners have, since the minimum is 20.19 and Node 20 itself passed end of life in April 2026; whether your CI can set the API key environment variable rather than logging in, since that is the documented path; and whether the flows you run resolve their dependencies offline, because the CLI installs what each flow needs on first run.

Frequently asked questions

How do I install the QA Wolf CLI?

Globally with npm, pnpm, yarn or bun, using the scoped package name, or with no install at all by running the package through npx. Each GitHub release also carries precompiled binaries for Linux, macOS and Windows, which need no Node runtime. Node 20.19 or newer is required when you use the package, and the README notes that Node 20 passed end of life on 30 April 2026 and recommends Node 22.

What does the QA Wolf CLI do locally?

It runs and manages flows locally from a per environment cache directory, pulling the flows first when they are not already cached and installing the runtime dependencies they need. Flow creation, AI powered test generation, managed cloud execution and team collaboration are described as platform features rather than CLI features.

How do I authenticate the QA Wolf CLI?

With an interactive login command, or by setting an API key environment variable, which is the documented approach for CI. You need a QA Wolf account, and the environment identifier used by the flow commands comes from the dashboard under Settings, then Environments.

What is the QA Wolf runner SDK?

A TypeScript entry point for driving interactive runners without spawning commands, where every verb names the runner it addresses so nothing is launched or billed implicitly. A failed request returns a false flag covering transport, authentication and local validation, while a runner that refuses returns a success with an outcome and a failure reason from a published contract.

How do I record video of a flow in the QA Wolf runner SDK?

Generate a recording identifier yourself and pass it with the start action, keeping that identifier when retrying a start or a stop. The same verb accepts a status action and an automatic action, and a separate listing call takes optional recording and page filters, returning one page with stable platform URLs and expiring video URLs. Equivalent command line commands exist under the runner group.

Does the QA Wolf CLI ship an agent skill?

Yes. The package ships an agent skill that follows an open agent skills specification, installable globally or for the current project through a cross harness installer that detects supported harnesses. It is generated from a template and the command tree, kept in sync by the test suite, and it also ships inside the npm package for consumers who manage skill files themselves.

Official sources

  1. License: Apache-2.0
  2. Project website
  3. qawolf/cli on GitHub
  4. README
  5. Releases
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/qawolf-cli.svg)](https://hysenlabs.com/projects/qawolf-cli)

Star history chart for your README

markdown
[![Star history](https://hysenlabs.com/chart/qawolf-cli.svg)](https://hysenlabs.com/projects/qawolf-cli)

Was this useful?