Your app looked fine yesterday. Then a designer nudged a button, renamed a screen, and by this morning half your automated mobile tests are red. Nothing is actually broken in the app. The tests just can't find the elements they were told to click. If you have ever spent a Tuesday fixing tests instead of writing them, Loadmill is trying to hand that Tuesday back to you.
Quick facts
- Pricing (as of 3 Jul 2026): quote-based. The website points you to Request pricing and Book a demo rather than listing tiers, and it is aimed at teams and enterprises. The new Mobile Testing Agent desktop app is a free download while you evaluate it.
- Platforms: desktop app for macOS and Windows, testing both mobile and web apps. Runs headlessly in CI when you are ready.
- Open source: no. It is a commercial product, with ISO 27001 and SOC 2 certifications listed on the site.
What it is
Loadmill is a test automation company, and its newest piece is built around what the industry calls a computer-using agent, or CUA. That is an AI that operates software the way a person does: it looks at the screen, decides what to tap, and taps it. Instead of writing code that says "find the element with ID login_button_v2 and click it," you write "log in with a QA user," and the agent figures out the rest by looking at the app.
The important word there is looking. Traditional mobile automation leans on selectors, which are the internal names and coordinates of on-screen elements. When developers rename or move those elements, the selectors break, and the tests break with them. An agent that reads the visible screen doesn't care what the login button is called under the hood. It just sees a login button.
The real-world problem it solves
Picture a QA engineer at a food-delivery app. Every sprint, the design team ships small changes: a relocated cart icon, a reworded checkout button, a new step in the address form. None of it changes what the app does, but all of it changes the internal wiring the old tests depended on. So the engineer opens Monday with two dozen failing tests, spends the morning confirming that the app is actually fine, and patches selectors until the dashboard turns green again. Real coverage never grows. The team just keeps the existing tests alive.
Loadmill's pitch is to move the fragile part out of the equation. You describe the intent of a test in plain English, the agent performs it against a running app or simulator, and it hands back a reusable script you can review, edit, and rerun. When the UI shifts, the agent adapts to what it sees rather than snapping on a renamed selector.
Where it's useful
A QA lead at a fintech startup turns a Jira story like "user resets password from the login screen" straight into a runnable test, instead of assigning it to an engineer for half a day of scripting.
A mobile developer at a dating app running an iPhone simulator writes: open the app, log in as a QA user, swipe through five profiles, send a like, and confirm the match screen appears. The agent walks that flow and saves it, so the next release checks itself.
A release manager at a large telecom wires Loadmill into the existing CI pipeline so that every build runs a batch of natural-language flows headlessly and reports failures back in the console, no separate test rig required.
A product manager writing acceptance criteria hands the agent an epic in the same words the team already uses to describe features, and gets executable checks out the other side, closing the usual gap between "what we asked for" and "what we tested."
Why it stands out
The genuinely interesting bit is that Loadmill treats your test intent as the source of truth, not a pile of selectors. You already know what a test should prove. Getting that idea into reliable automation has always been the slow, brittle part, and that is exactly the part the agent takes over.
It is also refreshingly concrete about the format. Here is the kind of instruction the tool actually runs, straight from a mobile flow:
Open the app on an iPhone 15 simulator
Log in with a QA user
Tap "Start order"
Search for "pepperoni pizza"
Add the first result to cart
Verify the cart badge shows "1 item"
Proceed to checkout
Tap "Place order"
That reads like a note you would leave a teammate, and that is the point. A non-technical stakeholder can read the test and know what it checks.
We will add the honest caveat: an agent that interprets screens is doing more thinking per step than a dumb selector, so you are trading some raw speed and predictability for resilience. And because pricing is quote-only, you cannot size the cost from the website. For a small side project that math may not land. For a team drowning in test maintenance, it very well might.
How to try it
The quickest path is the Mobile Testing Agent desktop app. Download it for macOS or Windows from the site, connect a simulator, and describe a flow you already test by hand. You should see your first AI-generated mobile test within a few minutes.
If you want the fuller platform, which also covers API and web UI testing, the route is booking a demo and requesting pricing, since those plans are sold to teams rather than self-served.
Takeaway
Loadmill is betting that the future of mobile testing looks less like maintaining brittle scripts and more like writing down what you want checked and letting an agent go check it. If your team spends more time repairing tests than adding them, the free desktop app is a low-stakes way to find out whether that bet pays off for you.