Check out a cart

Explore a cart and checkout before a release, read the bug it finds and keep tests of the main journeys.

This example explores the cart and checkout of Parcel at https://parcel.example before a release. You ask for it in one message and approve a plan, and subagents work through the cart, the checkout and the saved cards while signed in as your test account. You get the bugs they found, then tests of the main journeys that you can run before every release.

Before you start Link to Before you start

A cart belongs to someone who has signed in, so add a test account for the subagents: in Settings, open Test accounts under the project and add a Password account named member. The exploration uses the project's test accounts by itself, and if there is none when it meets the sign-in page, Re:agent asks how to get in and you can add one there.

Ask Re:agent Link to Ask Re:agent

Open Re:agent, start a New chat and say what to check and what to leave alone:

Output
Explore the cart and checkout before the release. Leave the team settings alone.

Re:agent starts an exploration from that message, with your first sentence as what to check and your second as what to leave alone. Explore your app shows the other ways to start one.

Approve the plan Link to Approve the plan

Re:agent first goes through Parcel on its own, then writes the plan as one card in the chat, Exploration plan. For this message it holds parts such as Cart, Checkout and Saved cards, each with its checks and a switch. Under them, What the plan won't do starts with Pay for anything and ends with your own words, Leave the team settings alone.

Read it, turn off any part you do not want, and press Start testing.

What the subagents do Link to What the subagents do

Subagents start on the parts, each in a browser of its own, and the panel beside the chat lists their test cases as they go, grouped by part and then by page. They add boxes to the cart and change quantities, and whatever the plan said it adds, such as a saved card, is really added in Parcel. When a subagent reaches Place order, it stops there, because an exploration never pays for anything. It checks everything up to the order and never places one.

When a check fails, a second subagent repeats it in a fresh browser, and only a failure that happens again becomes a bug.

Read the bug Link to Read the bug

When the exploration ends, the chat says Exploration complete. Here's the report. The report's Bug report lists Bug #5, Saved card disappears after signing in. Press it to open the bug: the recording with the bug's steps marked on it, then Steps to reproduce, Expected, "The card saved before signing out is offered at checkout", and What happened, "Checkout asks for a new card".

From here, Copy prompt for your coding agent hands the bug to Claude Code, Codex or Cursor, and Fix a bug with your coding agent follows a bug like this one to a checked fix.

Create the tests Link to Create the tests

Under Test report in the same card, press Review the test cases. The panel opens on a card that recommends your main journeys, the parts where people make an account or save something, such as Cart and Saved cards, with how many tests each makes. Press Create tests from the main journeys, or Choose parts yourself to pick others, such as Checkout.

Making tests is free, and Rehearsal makes up to 20 at a time. Each test is written from what a subagent saw work and runs twice, the second time in a fresh browser, while the card says Checking 6 tests and then 6 tests ready.

Run them Link to Run them

Press Run once now on the same card. The run dialog says what the run uses, such as "This run checks 6 tests on 1 screen: 6 test runs.", and Run starts it. The tests stay under Tests from then on, so you can put them in the release suite and run them before every release, or let pull requests run them.

Next steps Link to Next steps