Skip to content

Write a QA journey, run it, read the verdict

In this tutorial we will write a small Agentic QA journey for your own website, run it from the terminal, and read the report it prints. Along the way we will make one step fail on purpose, so you have seen both a passing step and a Finding before you write a real regression suite.

You do not need to know how Agentic QA works inside. We will write two steps, run them, and look at what came back.

You need the CLI, a signed-in session, and a Binclusive project whose website is set up in the dashboard:

Terminal window
npm i -g @binclusive/cli
b8e auth login
b8e project

b8e project lets you pick the project we will work in. The journey will run against that project’s website. The project needs Agentic QA on its plan; when it has it, the dashboard shows an Agentic QA section for it.

Open the website’s landing page in your browser and keep it open. We need three things from it:

  • the text of the landing page’s main heading;
  • the name of one link in the main navigation that opens another page of the same site, in the same tab;
  • the main heading of the page that link opens.

In this tutorial the landing page heading reads Welcome, the link is named Pricing, and the page it opens has the heading Pricing. Wherever you see those words below, use what your own site shows.

Create a file called journeys.json:

[
{
"name": "Landing page reaches pricing",
"startPath": "/",
"steps": [
{
"instruction": "Look at the main heading of the page",
"expectedResult": "The heading reads \"Welcome\""
},
{
"instruction": "Click the \"Pricing\" link in the main navigation",
"expectedResult": "The page's main heading reads \"Pricing\""
}
]
}
]

Each step is something a person would do, written as a person would say it, plus what they should see afterwards. startPath is where the journey opens: / is the landing page of the project’s website.

Terminal window
b8e qa journeys import journeys.json

The CLI checks the whole file before it creates anything, and then prints the new journey with its id. List the project’s journeys to see it there:

Terminal window
b8e qa journeys list

Run the journey by its name:

Terminal window
b8e qa audit --journeys "Landing page reaches pricing" --no-headless

--no-headless opens the browser window, so you can watch it work. A browser opens on your landing page, reads the heading, then clicks Pricing. In the terminal, each step prints a label as it finishes. Both steps should print PASSED, and the command ends with the report.

Check the exit code:

Terminal window
echo $?

It is 0: every step passed.

Now we will see what a Finding looks like. Open journeys.json and change the first expected result to a heading your page does not have:

"expectedResult": "The heading reads \"Goodbye\""

Import the file again:

Terminal window
b8e qa journeys import journeys.json

Importing a file twice creates its journeys twice, so the project now has two journeys with the same name. Run b8e qa journeys list and copy the id of the newest one, then run it by id:

Terminal window
b8e qa audit --journeys qaj_your_new_id

This time the first step prints FAILED. The report says what the page showed instead of “Goodbye”. The second step still runs and passes, because one failed step does not stop the journey. The exit code is now 1: the run has a Finding.

Every run is stored. List them:

Terminal window
b8e qa audit list

Copy the id of the failed run and print its report again:

Terminal window
b8e qa audit report qar_your_run_id

It is the same report the live run printed. The same run is also under Agentic QA → Runs in the dashboard.

Delete the journey with the wrong expectation:

Terminal window
b8e qa journeys delete qaj_your_new_id

Its runs and Finding stay in the project’s history.

We wrote a journey as plain steps and expectations, imported it, ran it in a real browser, and saw the two verdicts that matter: a step that passed and a step that failed. We also saw that the exit code follows the verdict, which is what lets a script or a CI job act on a run.

Next: