Question 1

What is Thunder by TestMu AI?

Accepted Answer

Thunder is TestMu AI's Chrome extension for authoring load tests, currently in beta. It writes an Apache JMeter .jmx, or the same plan as a k6 script, from a recording or a file your team already keeps, and it can trigger the run on HyperExecute.

Question 2

Is Thunder free?

Accepted Answer

Yes. Thunder is free while it is in beta, and it needs a TestMu AI account, which is also free. Runs on Performance Testing by HyperExecute draw on your account's plan. The Free plan includes up to 100 concurrent virtual users and 100 virtual user hours a month as an introductory offer, and the performance testing pricing page lists every tier.

Question 3

Why does Thunder need Chrome's debugger permission?

Accepted Answer

To read response bodies, which correlation depends on. The DevTools protocol is the only Chrome extension API that exposes them; the webRequest API sees requests but not their response bodies. Thunder attaches only to a tab you start recording.

Question 4

Why doesn't Thunder run in Firefox or Safari?

Accepted Answer

Firefox does not implement the chrome.debugger API, so an extension there cannot capture response bodies, and Thunder cannot correlate without them. Safari is not supported either. Use a Chromium-based browser instead.

Question 5

Why does my JMeter plan pass with one user and fail under load?

Accepted Answer

Usually because a recorded value was replayed after it expired, or a login token was stored where other threads cannot read it. The JMeter user manual explains that variables are local to a thread while properties are common to all threads. Thunder's Authentication option shares a single login across every user, so for a login per virtual user, keep the login inside the recorded journey and let each thread correlate its own token.

Question 6

Which sources give Thunder something to correlate?

Accepted Answer

Recordings, and TestMu AI sessions captured with full HAR, because only they carry response bodies. Plans from cURL, an OpenAPI spec, a spreadsheet or a URL list start with no correlated values, and a Postman collection brings only the chaining its scripts already define. Those plans still serve API load testing from the first request.

Question 7

Can I turn a Selenium test into a load test?

Accepted Answer

Yes, if the session ran on TestMu AI with network.full.har set to true in its capabilities. Paste the session ID, or pick from your 40 most recent sessions, and the plan comes back grouped into the steps the test took, such as Open, Click and Type into. Playwright sessions are not supported yet. For the functional suite itself, see Selenium automation.

Question 8

Does Thunder keep the think times from my recording?

Accepted Answer

Yes, by default. Each request is followed by a timer for the gap that actually occurred while you browsed, so virtual users pace like people. Turn it off and every user replays the journey as fast as the server answers, which is a stress test rather than a load test. Randomizing spreads each pause between half and one and a half times the recorded gap, so users that started together drift apart.

Question 9

What if the journey spans two of our domains?

Accepted Answer

Tick both under Hosts in this recording and rebuild the plan. Thunder keeps the busiest domain that is not a tracker, with every subdomain of it, so app, api and auth subdomains stay without any ticking. A login service on a separate domain needs the tick, which also stops the third-party filter from dropping it.

Question 10

Can Thunder check a .jmx file I already have?

Accepted Answer

Yes. Validate .jmx checks the file's encoding, characters XML does not allow and unclosed tags, and confirms it holds a test plan, a thread group and a sampler. Each problem is reported by line and column, and a 37 MB plan is checked in under a second. The check is strict on purpose: HyperExecute's parser rejects a plan carrying control characters even when JMeter itself still opens it.

Question 11

How does a k6 script from Thunder fail a run?

Accepted Answer

Through thresholds. The k6 documentation on checks states that failed checks do not cause a test to finish with a failed status, so every generated script requires checks to pass above 99% and failed requests to stay below 1%, with an optional p95 budget. Crossing one exits the run with code 99. k6 testing on HyperExecute covers the job side.

Question 12

Does Thunder support client certificates and mTLS?

Accepted Answer

Yes. Keystores are set while the plan is built, and Thunder generates and checks the JVM properties they need at that point instead of leaving a failure for run time. Upload the keystore and its system.properties file with the plan.

Question 13

Can Thunder model arrival-rate load?

Accepted Answer

Yes, in the spec editor. An arrival-rate thread group keeps starting new users at a set rate while responses slow down, where an ordinary thread group waits for each response before sending more. The runner needs the jmeter-plugins-casutg jar, so upload it with the plan. The load testing guide covers when each model fits.

Question 14

Can I save the raw recording?

Accepted Answer

Yes. Export HAR saves the recording with your transactions, assertions and extractors inside it, and the file stays a valid HAR that DevTools and other tools read. To look through one in the browser, open it in the free HAR file viewer.

Question 15

What happens to our plans if we stop using HyperExecute?

Accepted Answer

They keep working. Thunder writes plain Apache JMeter, so a .jmx runs on a laptop, in Jenkins or on any platform that runs JMeter. On HyperExecute a plan runs as a JMeter project, and JMeter load testing in the cloud covers that path.

Question 16

What can't Thunder do yet?

Accepted Answer

It has no trend charts or run-to-run comparison yet, and that is first on our list. A pasted cURL command keeps its own Authorization header, which can override a fresh token, so remove it before generating. Removing a wrong low-confidence correlation currently means editing the plan's spec.

Question 17

How did TestMu AI run the Thunder benchmark?

Accepted Answer

We built a recording with ten planted cases: eight values that must be correlated, a session cookie that belongs to the cookie manager, and one repeated value that is not dynamic. A value counted as correlated only when its recorded literal was gone from every later request and an extractor in the plan found it. Both plans then ran in JMeter 5.6.3 against a local server that rejects stale values. The results describe one free converter, not every tool its vendor makes.

Thunder: JMeter Recorder and Script Generator

How Thunder Builds a Load Test

Start From the Files You Already Have

Record the Journey in Your Own Chrome

See Every Token Thunder Wired In

Prove the Plan Works at One User

Keep the Recording in Your Browser

Send the Plan to HyperExecute

Built for Plans That Hold Up Under Load

Thunder vs a HAR Converter

Pick the Output Your Test Needs

Install Thunder and Build a First Plan

Add It to Chrome

Pin It to the Toolbar

Sign In to TestMu AI

Generate a First Plan

More Reasons to Love TestMu AI

TestMu AI positioned as a Leader in the 2026 Gartner® Magic Quadrant™

TestMu AI recognized in The Forrester Wave™: Autonomous Testing Platforms, Q4 2025

Wall of Fame

Enterprise-Grade Security

Integrations

Frequently Asked Questions