AI builder
From a sentence to a working workload.
Brief Studio the way you would brief a colleague. It works on the real application, so what it builds is what your users click.
LoadGen Studio 2026 runs on Windows and macOS. It is in early access for LoadGen customers: ask LoadGen Support to join, then download it from the software updates page of your LoadGen Platform.
- Web
- Citrix
- Citrix Enhanced
- RDP
- AVD
- Omnissa Horizon
- Local Windows apps
- Mobile
- 01
Describe
Type the journey in plain language, with the data and the timings you care about.
- 02
Plan
Studio splits it into blocks, each with a start state and an end state.
- 03
Build every screen live
It drives the app screen by screen and writes validated steps as it goes.
- 04· optional
Replay & heal
It replays the result with the full pipeline and repairs what broke.
Recent
What to do
How should it run?
Approve, edit or skip every step before it happens.
AI sees
Screenshots and the page structure are sent to the configured AI provider during a build. Nothing is saved to the workload without your action.
Human in the loop
You decide what the AI may do on its own: Ask first or Auto per technology, and for each guardrail whether it asks, allows or refuses. Each Studio user controls these settings, and every guardrail starts at Ask.
Pick "Ask first" and approve, edit or skip every step. Pick "Auto" for the parts you trust. Guardrails decide what always needs you.
- Auto
- Runs on its own, and stops where you set a guardrail or an action to ask.
- Ask first
- Approve, edit or skip every step before it happens.
- A Submit, Send or Delete click the task never asked for
- Leaving the allowed site or application
- Creating a variable the task never asked for
- A third attempt on a target that missed twice
- A final tidy-up that deletes many steps
The Ask, Allow and Refuse values above are an example. Each Studio user controls these settings, and every guardrail starts at Ask. Change one and the session on the right behaves accordingly.
Action
Click "New claim" (mark 1)
Why
The task starts a new travel claim.
Load-readiness review
A green replay is not yet a good load test.
Before you keep a take, Studio checks it the way a load tester would: no shared accounts, unique data per virtual user, real think time, sync points and named measurements.
Where a check has a fix, apply it, let the AI fix it or add a cleanup block. Otherwise accept the finding as is, or ignore it.
How LoadGen runs load testsKeeping a take and opening it in the editor happens in LoadGen Studio. This page only demonstrates the review.
Hard-coded account
"jdoe" typed in block 1, Sign in
Unique data per user and iteration
Every virtual user files the same claim title
Data that grows every iteration
Each iteration leaves a submitted claim behind
- Secret typed as plain text
- Think time
- Every interaction waits on a sync point
- Measurements named, unique and stopped
- Clean start and end
- Coordinate clicks
- Unchecked key presses
7 of 10 checks pass.
Watch it think
Watch it think, and steer it.
While Studio builds, it shows what it is doing now and why, block by block, and you can export the plan and its reasoning. Tell the AI what to change, take over the next steps yourself and hand back, or pause it. Repairs arrive as drafts you keep or undo.
It shows its reasoning
What it is doing now and why, block by block, and the plan and its reasoning export with it.
You can take the keys
Tell the AI what to change, take over the next steps yourself and hand back, or pause it.
Repairs are drafts
A repair arrives as a draft you keep or undo, not a silent change to the workload.
Self-healing
When a step breaks, it shows you what it changed.
Heal a single block, or fix a step with the AI and run to that point to confirm it. The repair is a change you can read and accept, not a silent retry.
A workload you built once moves with you: to a new gold image, or to a desktop in another language. Open each block in LoadGen Studio against a profile on the new desktop and choose Heal. The AI replays the block, works out why a step failed and repairs it, and you keep or undo the repair. Heal an English workload on a Spanish desktop, block by block, and run it without recording it again.
Heal this block
Repair one block and leave the rest of the take alone.
Fix with AI and run to here
Apply a suggested fix, then replay up to that step to see it hold.
Optional full replay check
Stage 4 replays the whole take with the full pipeline and repairs what broke. It is optional, and off unless you ask for it.
Privacy and choice
Pick who runs the model. Keep your secrets out of it.
Nothing to set up
LoadGen AI credits
LoadGen Studio uses the LoadGen Cloud account it already signs in with, and the AI is paid from your organisation’s prepaid credits: one balance, shared by everyone.
- Choose a model tier and an effort for the build
- See the credits each build and each block used
Bring your own
Your own API key
Use your own account with a supported AI provider. The key is stored encrypted on your machine.
- Pick the model and the effort
- No LoadGen AI credits needed
On-prem
Local or custom server
Point Studio at a model server on your own network. Nothing leaves your network.
- Fits air-gapped and regulated sites
- Same builder, same guardrails
- Off until you switch it on
- Secret values are never shown to the AI
- You choose what it sees
- Allowlists for sites and apps
- Every step shown, with its reasoning
Screenshots and the page structure are sent to the configured AI provider during a build. Nothing is saved to the workload without your action.
Migrating
Bring your Katalon tests with you.
Moving from Katalon Studio? Import a project with its profile variables, replay the tests in LoadGen Studio 2026, and let the AI Builder repair the steps that fail, asking first or on its own. In Functional mode, describe a new testcase and the AI builds it, or heal a failing testblock and keep or undo the change.
Brief it on an application you actually run.
We describe one of your journeys in plain language, let Studio build it on the real application, and walk through the load-readiness review with you.
