AI Studio

AI Studio is the authoring workspace for AI-generated automation. You describe what to test, an AI agent drives the application to work out how, and the result becomes an executable scenario.

Open it from AI Studio in the header of Scenarios or Test Plans. The first time you do, the workspace's AI source is provisioned automatically.

AI Studio replaces the normal shell chrome with its own header, which carries a source switcher and a Back to AutoRunner button.

note

AI Studio is in beta. Features and behavior may change.

Tabs

A narrow icon rail on the left switches between these areas.

Tab	Purpose
Scenarios	The scenario tree and the scenario detail panel
Test Files	Test data files. Shown only for mobile sources.
Element Repository	The registry of screens and elements
Screens	Screen knowledge, screenshots, and application flows
Custom Steps	Reusable composite steps
Settings	Generation defaults for this source

The Scenarios icon shows a pulsing badge with the number of active generation runs.

Element Repository and Screens are documented in Element Registry.

Scenario Tree

The left panel of the Scenarios tab is the scenario tree.

• Search scenarios filters as you type.
• New Folder, Select All / Deselect All, and Refresh sit in the tree header.
• New Scenario is at the bottom.

Filters

Three multi-select dropdowns, each with a Clear all entry. Active values appear as removable chips.

Filter	Options
Generated	Generated, Not generated
Status	Generating, Pending, Pass, Fail, Error, Aborted
Platform	Web, Android, iOS, Desktop

note

A generation in progress shows as Generating rather than "Running", which is reserved for test runs.

Bulk Actions

Checking scenarios reveals a bar with Generate, Move, and Delete. Scenarios already generating are excluded, and the bar explains that you need to deselect them first.

Row Menus

Folders offer New Subfolder, Rename, Move to..., Duplicate, and Delete. Scenarios offer Rename, Duplicate, Move to..., Copy Link, and Delete.

Creating a Scenario

New Scenario offers two paths:

• Import from Test Repository — bring in existing test cases.
• New Scenario — write one from scratch.

The scenario form has:

Field	Notes
Scenario Name	Required
Folder	Optional; you can create a folder inline
Pre-Conditions	What must already be true before the scenario runs — a logged-in user, seeded data, a specific page

Then choose how you describe the test:

• AI Prompt — describe the test in natural language. Type $ to insert a variable.
• Step by Step — enter each step's action and expected result, reordering by drag.

Finish with Create Scenario, or Create & Generate to start generation immediately.

Generating Automation

Generate in the detail panel header starts a generation run. Once steps exist, the button becomes Re-Generate. Stop ends a running generation.

Save any pending edits first — generation is blocked while there are unsaved changes.

Choosing a Target

The generate dialog first asks for a platform:

Option	For
Web	Browser-based tests
Android	An app or a browser on an Android device
iOS	An app or a browser on an iOS device
Quick Run	Uses your saved defaults. Disabled until you configure them under Settings.

When the scenario already has generated steps, Generate into a new scenario copies it first so the current scenario keeps its steps.

Web

Maintain a list of Target URLs and pick which one is the default. The default is used when a scenario has no start URL of its own.

Mobile

Choose the Test type:

• Native App — install or launch an app on the device.
• Mobile Browser — open the scenario in a browser on the device.

Then choose the device. Auto-assign lets the system pick an available device, or Select specific device opens the device list with a search box, a Device Type filter (Real Device or Emulator), and a version filter. Devices show as Online, Available, Busy, or Offline. Choosing a busy device turns the action into Queue Run — the run queues and starts when the device frees up.

For a native app, pick the application from one of four sources:

Source	Content
Installed	Apps already on the device
Distribution	Firebase App Distribution releases
Files	Previously uploaded builds
Upload	A new APK, IPA, or AAB

See Devicer for the device farm itself.

Regenerating from a Step

Any step's menu offers Regenerate from here.

Steps before the chosen one are replayed deterministically in one session, then generation replaces that step and everything after it. You can review or edit the generation prompt first, and tick Create a new scenario to write the result into a copy instead of overwriting.

warning

Regenerating from a step inside an IF, FOR, or WHILE block affects the whole block.

Resuming a Generation

A generation run that stopped part-way shows a Resume action in the History tab, which continues from the stored checkpoint rather than starting over.

Scenario Detail Panel

The right panel has seven tabs.

Tab	Content
Scenario	Pre-conditions and either the AI prompt or the manual steps
Script	The generated automation steps and their editor
Live	The live view of a running generation. Enabled only while a run is live.
Variables	Scenario variables
Versions	Automation versions and activation history
History	Generation runs for this scenario
Files	Test data files for this scenario

The header also carries Link to TestPilot, Share Link, Stop, and the generate button.

Script

The Script tab holds the executable steps.

Toolbar actions: View Generation, Search (Ctrl+Shift+F), Timeline, Variables, and Edit / Save.

Timeline opens an animated walkthrough of the run's screens, with play, pause, step navigation, and 1x to 8x speed.

In edit mode you can:

• Select steps from the left gutter and Wrap Selected Steps in an IF, WHILE, or FOR block, then set the condition.
• Use the step bulk bar for Update Settings, Wrap in Block, Copy, Create Custom Step, and Delete.
• Use a step's menu for Insert step below / above, Move up / down, Duplicate, Copy step text, Run only if a condition…, Settings, and Delete.

Elements referenced by a step are picked from the Element Registry, which is what makes a generated scenario survive UI changes.

Versions

Automation is versioned. Each version is one of:

State	Meaning
Active	Runs use this version. Editing it creates a draft first.
Review draft	Saved but not active. Review it, edit if needed, then publish explicitly.
History	Read-only. You can preview it or make it active again without creating a new version.

Actions include Edit draft, Save draft, Publish, Discard, Preview, Make active, and Delete version, plus bulk delete of selected versions. Publishing validates and compiles the draft; future runs use it only once both succeed.

Rows carry warnings where they apply — the document schema version, steps that will not run, steps that need review, and validation issues.

Below the version list, Activation history records publish, restore, deactivation, and import events separately from the versions themselves.

History

Every generation run for the scenario, with its status and result, plus Resume where the run can be continued and View Agent Conversation for a read-only replay of what the agent did.

Working with a Running Agent

While a generation is live, the Live tab streams the browser or device.

An activity bar over the viewer names what the agent is doing — Starting, Understanding, Planning, Acting, Reviewing, Thinking, Paused, Waiting for you, Finished — with a button that opens the agent panel.

Conversation

The RabbitQA Agent panel shows plans, actions, blockers, and your guidance as one timeline. Messages are labelled You, System, or Agent.

When the agent needs input, it posts a Waiting for your answer card and the panel shows a Needs answer pill. Type your answer and send it to unblock the run.

Pause and Resume

A single Pause / Resume button controls the agent, with PAUSED and WAITING FOR INPUT status badges. While paused you can queue a message for when the run resumes.

Pause is disabled while a question is pending — answer it first.

To end the run entirely, use Stop.

Custom Steps

A custom step is a named BDD pattern with typed parameters that expands into an ordered list of existing steps — a reusable login flow, for example.

Field	Notes
Step Key	Lowercase snake_case, no spaces. Fixed after creation.
Platform	Android, iOS, Web, or Desktop
BDD Pattern	Use <param> placeholders, for example Login with "<username>" and "<password>"
Folder	Defaults to Unfiled
Parameters	Derived automatically from the placeholders in the pattern; set each one's type and display options
Child steps	The steps the custom step expands to, reorderable by drag

Child steps bind their elements by qualified name — Screen.element — drawing on the active registry elements across every platform.

You can also create a custom step from a selection in the Script tab with Create Custom Step.

Settings

Generation defaults for this source, in four sub-tabs.

AI Config

Assertion Level controls how many verification checks the AI generates:

Level	Behavior
None	No assertions — actions only
Minimal	Only critical checks, such as page loads and form submits
Moderate	Balanced — UI elements and data validation. The default.
Comprehensive	Detailed checks on every step

AI Generation Instructions holds project-level rules applied to every web and mobile generation — for example, that a cookie popup is optional, that payment confirmation steps must never be skipped, or which test account to use for login.

Scenario Steps

• Setup Steps (Before) run before each scenario starts, such as logging in or navigating to a starting page.
• Cleanup Steps (After) run after each scenario ends, such as logging out or clearing data.

Both lists are reorderable by drag.

Web

Web Defaults for AI web runs:

• Saved URLs — the entry points, with one marked default. The default is used when a scenario has no start URL.
• Proxy — a proxy server and a bypass list.
• Chrome Hub tags — route web runs to matching Chrome Hub nodes. Tags are lowercased and de-duplicated.

Mobile

Mobile Defaults for mobile generation: a Default Platform, a Default Device (auto-assign or a specific device), and a Default App. These are what Quick Run uses in the generate dialog.

Linking to CaseWriter

Link to TestPilot in the detail panel header creates a test case from the scenario and keeps it in sync.

Pick a Test set and a Priority, then Create & Link. Later edits to the scenario's name, description, steps, expected results, and preconditions sync to that case automatically.

Once linked, the control becomes a CaseWriter badge that opens the linked case in CaseWriter's test repository in a new tab.

note

Linking reads test sets and priorities from the test management module. If you do not have access to it, the dialog cannot load its options.

Next Steps

• Element Registry — Screens, elements, locators, and the proposal queue
• Test Plans — Put generated scenarios into a runnable plan
• Run Detail — Analyze what the generated scenario did
• AI Editor — The code editor for repository-backed sources