Create a Load Test with AI
Create with AI turns a plain-English description into a runnable JMeter or k6 load test. You describe what to test, review the test plan it builds, validate it with a real one-user run, then run it like any other test.
Open Create with AI
- In JMeter or k6 load testing, click Create with AI in the left menu.
- Or, on the new test page, click Describe it and let AI write it under the upload area.
1. Describe the test
Under What do you want to test?, write what you want in plain English, up to 4,000 characters. Be specific: the URL or endpoints, how many users, how long to ramp up and hold, and what counts as fast enough. For example: "Load test my API with 200 users ramping over 2 minutes, hold for 5 minutes, p95 under 1 second".
Then choose:
- Engine: Auto, JMeter or k6. Auto picks JMeter when your description mentions JMeter, and k6 otherwise.
- Add a cURL command (optional): paste a request, so the test uses its exact URL, method, headers and body.
- Import a recording or API spec (optional): a HAR recording, an OpenAPI file or a Postman collection. Two checkboxes add image, CSS and JS requests and third-party hosts. Both are off by default.
Or try an example fills in a sample description: a website, an API endpoint, a login and checkout journey, a spike test, a stress test or a cURL command.
Click Generate Test Plan. This uses 1 AI credit.
2. Review the test plan
The Test Plan screen shows what was generated. In the Visual view you can change:
- Users, Ramp up (s), Duration (s) and Regions,
- the Thresholds: p95 (ms), p99 (ms), Error rate (%) and Min RPS.
Scenario steps lists the requests in order.
The Code view shows the script: a k6 JavaScript file or a JMeter .jmx file, depending on the engine. Copy and Download take the script with you. Detach and edit code lets you change the script by hand. Click Save code to keep your changes.
Click Continue to Validate.
3. Validate with a smoke run
On Validate your test, LoadFocus first checks the plan and repairs the errors it can, shown under Automatic repair. When the checks pass, click Validate to start the Smoke run: a real test with 1 virtual user and 1 iteration, from US East (N. Virginia). It shows how each step did.
The smoke run is a real test run, so it counts toward your daily test limit like any other run. The note above the button tells you how many daily tests you have left. If your team already has a test running, or reached its daily, monthly or yearly test limit, the smoke run is skipped and says why.
Validation can also include an AI review of the plan, which uses 1 AI credit. Without credits the review is skipped and the other checks still run.
The verdict is one of:
- Passed: the test is ready to run.
- Failed: fix the issues listed before you run it at scale.
- Partial: the smoke run gave no clear pass or fail. Review the checks.
Run on LoadFocus and Download appear when the verdict is Passed or Partial.
Fix with AI
When a step fails in the smoke run, click Fix with AI next to it. The AI changes the plan, then you validate again. Each AI fix uses 1 AI credit.
Fixes are limited per plan: 2 when you click Fix with AI, up to 4 during automatic repair. After that you see "Auto-fix limit reached for this plan" and edit the plan yourself. Repairing code you changed by hand with Detach and edit code needs a paid plan.
4. Run the test
Click Run on LoadFocus. LoadFocus saves the test in your JMeter or k6 tests, starts it with the users, duration and regions of the plan, and opens its live results. The run counts toward your plan limits like any other test.
If the run cannot start, for example because you reached a limit, the test is still saved. Start it later from your JMeter or k6 tests list.
From then on it is a normal test: run it again, schedule it and analyze its results. See Run a k6 load test and Run an Apache JMeter load test.
AI credits
Create with AI uses the same AI credits as AI analyses of results, from the allowance for the engine you picked (JMeter or k6).
| Action | AI credits |
|---|---|
| Generate Test Plan | 1 |
| AI review during validation | 1 |
| Each AI fix, by hand or automatic | 1 |
| Static checks and the smoke run | None (the smoke run uses a test run instead) |
When you have no credits left, Generate Test Plan stops and says you are out of AI credits. If you can buy more, the message says so. If an AI step fails and produces nothing, a purchased credit it used is returned. See AI credits for when plan credits come back and how to buy a pack.