Autonomous AI QA agent

Give it a goal. Get a verified QA report.

Reliqa is an autonomous AI QA agent that tests web products in a real browser and returns a verified PASS/FAIL report with evidence.

Reliqa opens a real browser, explores your product, and checks whether the outcome actually happened. You get PASS/FAIL with steps, screenshots, and video — not a script that breaks the next time a button moves.

Built for teams who need evidence, not another brittle test file.

Multi-sense observationSeparate verifierEvidence-backed findings

Scripts test what you remembered to write.

Traditional E2E suites encode one happy path and break when the UI changes.

The locators break when the UI changes. Assertions miss the bug a real user hits. Exploratory testing does not scale, and a failed CI job often cannot tell a product defect from a bad selector.

Brittle scripts

A renamed button fails the suite and teaches you nothing about the product.

Blind spots

If nobody wrote the case, nobody tested the path.

Weak evidence

A red X without reproduction steps, screenshots, and network context wastes the next hour.

A browser agent that can explain itself.

Reliqa is a goal-driven browser agent with a separate verifier and evidence-backed findings.

Goal-driven runs

Write the outcome in plain language. Reliqa chooses the steps.

Outcome: Coverage without maintaining a script for every click.

Multi-sense observation

It reads the screen, the DOM, the accessibility tree, the console, and the network together.

Outcome: Fewer “it looked clicked” false passes.

Separate verification

Completing actions is not the same as meeting the goal. A verifier checks evidence before a pass.

Outcome: You can trust PASS, and FAIL says whether the app broke or the run was inconclusive.

Evidence-backed findings

Severity, reproducibility, steps, screenshots, video, console, and network.

Outcome: Bugs arrive ready to file.

See. Think. Act. Verify.

You give Reliqa a URL and an objective. It explores in a real browser, then verifies the outcome.

  1. 1

    See

    Reliqa observes the live page through multiple signals.

  2. 2

    Think

    A vision model decides the next action (click, type, wait, done, fail).

  3. 3

    Act

    Playwright performs the action in a real browser.

  4. 4

    Verify

    A separate check confirms the goal before a pass is accepted.

Multi-step flows

Chain goals such as login, then invite a teammate, then confirm they appear.

Site discovery

Map pages and interactive elements before a run.

Cached successful paths

Replay a known route, then fall back to the agent if the UI changed.

Live run view

Logs, screenshots, video, plus stop and re-run.

Test the path you know. Then try to break it.

Reliqa has three run modes: Standard, Exploratory, and Chaos.

Standard

Complete the objective and verify it.

Example: “Reset password and confirm the user can sign in.”

Exploratory

Search past the happy path. Report bugs with reproduction, not a pile of random clicks.

Example: “Wander settings and flag anything that loses data or blocks a save.”

Chaos

Messy input, repeated clicks, latency, and dropped requests. Use this when you want to know what fails under stress.

Example: “Rage-click submit while the network drops requests.”

Who Reliqa is for

Reliqa is best for QA leads, developers, and founders who need evidence-backed end-to-end and exploratory QA without maintaining brittle scripts.

QA lead

More exploratory depth without hiring another week of manual cycles.

Developer

A failing run that includes the step list and a video, not a selector autopsy.

Founder / PM

A way to ask “does this flow still work?” without waiting on a test rewrite.

Written tests vs. a goal-driven agent

Reliqa is an alternative to maintaining brittle Playwright or Cypress suites for end-to-end and exploratory QA. It does not replace unit tests.

Test intent

Traditional E2E suite
Hardcoded steps and assertions
Reliqa
Plain-language objective, adaptive steps

Observation

Traditional E2E suite
Only what the test author wired in
Reliqa
Screen, structure, accessibility, console, network

Outcome

Traditional E2E suite
Pass/fail follows the script
Reliqa
Separate verification and failure classification

When the UI changes

Traditional E2E suite
Selectors and waits need edits
Reliqa
Agent replans; cached paths replay when still valid

Report

Traditional E2E suite
Depends on the framework
Reliqa
Findings with severity, reproduction, and evidence

Questions, answered plainly.

What is Reliqa, and who is it best for?

Reliqa is an autonomous AI QA agent that tests web products in a real browser and returns a verified PASS/FAIL report with evidence. It is best for QA leads, developers, and founders who need end-to-end and exploratory coverage without maintaining brittle scripts.

Is Reliqa an alternative to Playwright or Cypress?

Reliqa is an alternative to maintaining brittle Playwright or Cypress suites for end-to-end and exploratory QA. It uses Playwright as the browser actuator. You do not write Playwright or Cypress scripts. It does not replace unit or integration tests.

Do I have to write Playwright scripts?

No. You write the objective. Reliqa drives the browser with Playwright. You do not maintain selectors or test files.

How does Reliqa verify a pass?

Completing actions is not enough. A separate verifier checks evidence against the objective before a pass is accepted. FAIL says whether the app broke or the run was inconclusive.

What happens if the agent gets stuck?

Loop detection and a retry budget stop runaway runs. Inconclusive is reported as inconclusive, not as a product pass.

Can Reliqa run multi-step flows?

Yes. Chain goals such as login, invite a teammate, then confirm they appear on the team page.

Does Reliqa integrate with CI/CD today?

Missions run in the Reliqa web app. A public CI/CD plugin is not available yet.

Does a pass mean the product is perfect?

No. A pass means this objective was verified on this run, with evidence. It does not replace code review or unit tests.

How do I get access, and is there public pricing?

Access is currently limited and there is no public price list. Request early access to join the private beta waitlist and we will follow up.

Put a real agent on the flow that matters.

Send a URL and the outcome you care about. Reliqa comes back with a verified report.

Early access is limited. Join the waitlist and tell us what you want to test — we review every submission. See our Privacy Policy.