Category
AI QA testing: agents that use your app like your users
AI QA testing used to mean scripts that an AI helps you write. Now agents can use the app themselves. Here is what that finds, what it misses, and how to run it on every build.
What is AI QA testing?
AI QA testing uses AI agents to check a web app by using it: each agent opens the app in a real browser, tries a task in plain words, and reports what broke and where it got stuck. With Mimiq, each agent also acts as a different one of your users, so the report covers confusing steps as well as errors, and the browser records failing and slow requests on the step they happened.
Why teams look for it
Teams ship faster than they can write tests. A scripted end-to-end suite only checks the paths someone thought of, breaks when the page changes, and never notices that the button text sends people the wrong way. Manual QA is thorough and slow. AI QA testing sits between them: agents that explore like users, on every change, with nothing to maintain.
- No scripts: a link and a task in plain words.
- Up to 20 agents at once, each a different user in a real browser.
- Errors on screen, dead buttons, failed and slow requests, and confusing steps in one report.
- Runs in the app, from your coding agent, or on every preview deploy.
- Honest limits: not a deterministic regression suite, no native apps.
How it works
- 01
Give it a link and a task
Paste a public address, such as a preview deployment or staging, and say in plain words what a user would try, like "Sign up and start using it". No test scripts and no selectors.
- 02
Pick how many agents
1, 3, 5, 10 or 20 agents, 5 by default. Each acts as a different one of your users and gets its own real browser for up to 40 steps. About 5 minutes for up to 3 agents, about 5 to 15 minutes for 5, and about 15 to 30 minutes for 20. Agents take turns on shared browsers, so it can take longer when other runs are going, and you can close the tab and come back.
- 03
Read what they hit
The report shows what broke, what the browser caught on your site (JavaScript errors, requests that failed or answered 4xx or 5xx, and requests slower than 3 seconds), where each agent got stuck or gave up and why in its own words, with the step and a screenshot, and the problems ranked by how many agents hit them, each with a suggested fix.
- 04
Fix it and run the same agents again
Rerun on the fixed build and see whether the problem is gone for everyone who hit it. Or let your coding agent do the loop through the MCP server, or let the GitHub Action do it on every preview deploy.
What agents find that scripts do not
A script checks that a selector exists and a request answers 200. An agent acting as a user notices that the "Continue" button is below the fold on a phone, that the company field rejects a normal website address, that the error message does not say what to fix, and that after three tries it gives up. Those are the problems that cost you users, and no assertion was ever written for them.
Mimiq records both sides. The agent says what it saw and thought at every step, with a screenshot. The browser records what the agent cannot see: uncaught JavaScript errors, requests to your site that failed or answered 4xx or 5xx, and requests slower than 3 seconds. The report ranks the problems by how many agents hit them.
Where scripted tests still win
If a flow must behave exactly the same on every build, such as a payment calculation or a permissions check, keep a deterministic test for it. Agents are not deterministic: two runs can take different paths. That is the point when you want to know how newcomers fare, and the wrong tool when you need a yes or no on one exact behavior.
Agents are simulated users. They find problems fast and on every build; they do not forecast how many real people will convert, they are not a deterministic regression suite, and they do not install native mobile apps.
Run it where you already ship
In the app, paste a link. From Claude Code, Cursor, Codex, VS Code or Windsurf, connect the MCP server at https://mcp.mimiqai.com/mcp and your coding agent runs it on the preview it just built. On every pull request, the GitHub Action at github.com/victorgulchenko/mimiq-preview-check runs the agents on each Vercel, Netlify or Cloudflare Pages preview and keeps one comment up to date.
Questions
Do I need to write test cases?
No. You give a link and a task in plain words. The agents decide what to click and type, the way a user would.
Can agents get past a login?
With paid credits, agents can sign in with a test account you give them and open protected previews (a Vercel bypass secret or a site password). The secrets are filled in only in the browser and deleted when the run ends. Test accounts and protected previews come with paid credits: the $9 Starter or a plan from $19 a month.
What does it cost?
Your first test is free with a free account: sign in with Google and it starts right away, no card. The account includes 50 credits, enough for 10 agents. After that an agent uses 5 credits, and then it is a $9 Starter (120 credits that never expire) or monthly plans from $19.
Does it replace my end-to-end suite?
No. Keep deterministic tests for behavior that must never change. Add agents for what scripts cannot see: confusion, dead ends and problems on paths nobody wrote a test for.
Keep reading
Let the agents find what breaks first.
Your first test is free: sign in with Google and it starts right away. Last checked 2026-10-22.