Autonomous QA for web apps
Tell an agent what to test.
Describe what to check, pick an environment and a test account, and the Pervaro agent plans its tasks and works through your site in a real browser. Every finding comes with what was expected, what happened and a screenshot. No QA team required.
New organizations start with 100 free credits. Already have an account? Sign in
Task 1: Create an account
- Opened /pricing
- Clicked “Start free trial”
- Typed into the email field
- Read the confirmation message
The gap
Building got fast. Checking didn’t.
AI coding tools can produce an app in a day. Someone independent still has to prove it behaves as intended before users see it.
“It works on my machine” is the whole test plan.
The agent tests the deployed site the way a person would, in a real browser, against what you ask it to check.
A failing check might be a real bug, or just a flaky one.
Every finding shows what was expected, what happened and a screenshot, so you can confirm it or dismiss it.
Checking by hand takes time you don’t have.
Describe the goal in plain words. The agent plans its tasks, works through them, and asks you when it needs a decision.
What you get
An agent that tests like a person
From your instruction to findings you can act on.
Starts from a chat
Say what to test, then pick the environment and the test account to sign in as. Limits have sensible defaults, including a spend cap.
No test scripts to write before you begin.
Runs in a real browser
The agent plans its tasks, then clicks through your site in a real browser while you follow along.
You see what it does as it does it.
Signs in as your test users
Give the agent a test account. The platform types the password for it, so the agent never sees it and it is never shown again.
Testing behind a login, without sharing credentials.
Knows your product
Upload a brief, expected behaviour or test data, or paste it in. The agent looks things up in it while it tests.
Checks follow your intent, not just what the screen shows.
Asks when it is stuck
If it needs a decision it asks you. Blocked work is never counted as passed.
A check that could not finish is never reported as a pass.
Findings with evidence
Every finding shows what was expected, what happened and a screenshot. Confirm it or dismiss it.
Developers can reproduce a problem without asking you for details.
How it works
See Pervaro work
Step through a session below, from setting up an environment to reviewing findings.
Step 1 of 8
Start in Agent mode
New session is where you begin, and it needs an environment. Open Environments in the left menu to create one.
Create an environment first
An environment is the website the agent tests, with its test users and reference documents.
Create environmentStep 1 of 8: Start in Agent mode
Safety and control
Built to be pointed at your staging site
A tool that clicks through your product needs firm limits. These are enforced by the platform.
Guarded browsing
Targets pointing at localhost or internal network addresses are blocked before the browser opens them.
Write-only credentials
Test account passwords are encrypted at rest and never returned or displayed again.
You set the boundaries
Per-environment switches decide whether the agent may delete, pay, or send content. Delete and pay are off by default.
Roles for your team
Invite colleagues as Owner, Admin, Developer or Viewer. Viewers get read-only access.
FAQ
Questions, answered
- Pervaro is an autonomous QA agent for web applications. You describe what to check, point it at a test environment, and it tests your site in a real browser and reports what it finds, with screenshots.
- Teams that ship web software without a dedicated QA function, such as founder-led startups and teams building with AI coding tools. It's designed to take ownership of checking your site so developers don't have to.
- You create an environment (the site to test, with optional test users), start a session and say what to test. The agent plans its tasks, works through them in a real browser, asks you when it needs a decision, and reports findings with evidence.
- Create a workspace, add an environment with the site's URL and optional test accounts, then start a session and describe what to test. You can also upload a brief or test data for the agent to refer to.
- Each finding shows what was expected, what happened and a screenshot. You confirm it or dismiss it. Work that was blocked is never counted as passed.
- Test targets are checked so the browser can't be pointed at localhost or internal network addresses. Actions that delete data, spend money, or send and publish content are controlled by per-environment settings, and delete and pay actions are off by default. Even so, use a test environment and test accounts.
- Passwords you add for test personas are stored encrypted and are never returned by the API or shown again. Pervaro signs in on the agent's behalf, so the agent itself never sees them.
- New organizations start with 100 free credits, valid for 30 days. Credits pay for metered work such as agent sessions. Paid plan pricing isn't published on this page yet.
- Pervaro tests web applications through a browser. It can only test what it can reach, so staging access and working test accounts matter. Forms that require human verification, such as CAPTCHAs, aren't submitted, and password changes and account deletion are never automated. Teams can invite members with roles from Owner to read-only Viewer.
Put an agent to work on your site.
Create a workspace, add an environment, and start your first session.