NewTest design with coverage you can review

Tokay designs the tests. You approve them.

VibeQA turns a ticket into a test design your team can review: valid, invalid and boundary cases for every input. Then it automates the design and re-checks every fix in a fresh sandbox.

Free for one project · Web apps today, Android and iOS next

Tokay, the VibeQA gecko

To-kay. Every class has a row.

Storefront / CheckoutTest designCode and runsDraft · revision 9 · based on v12
3 AC · 8 cases · 25 rows+12 positive−13 negativeCoverage readyAsk TokayApprove v13
Requirements · SHOP-388
AC1 Delivery details are checked before payment+3−6✓ Covered
AC2 A customer pays by card or cash on delivery+5−3✓ Covered
AC3 A promo code reduces the order total+4−4✓ Covered
RowPhone numberDelivery addressLast run
10-digit mobile+84 format9 digits11 digitscontains lettersemptyHo Chi Minh Cityother province255 characters256 charactersempty
TC-01 · Valid delivery details continue to payment
TC-01.1✓ Passed
TC-01.2✓ Passed
TC-01.3✓ Passed
TC-02 · An invalid phone number is rejected
TC-02.1✓ Passed
TC-02.2✓ Passed
TC-02.3✓ Passed
TC-02.4✓ Passed
TC-03 · A missing or too long address is rejected
TC-03.1✓ Passed
TC-03.2New row
Covered41111151111
Comments
Comment on TC-03✓ Resolved

An Phạm · 12 min ago
@Tokay the address limit is 255 characters. Cover the boundary on both sides.

Tokay · 11 min ago
TC-01.3 already covers 255. I added TC-03.2 with 256 characters.

Tokay suggests 1 change+ TC-03.2 · address: 256 characters · error "Use at most 255 characters"Coverage afterAC1 empty classes 1 → 0✓ Accepted by Linh Nguyễn
Type @Tokay to ask for a change

Reads your specs and posts results where your team already works

JiraSlackTelegramNotionConfluenceGitHub

How it works

From a ticket to a suite you trust

Tokay does the writing. People make every decision that matters, and nothing goes live without one.

  1. 01

    Point Tokay at the spec

    A Jira ticket, a Confluence or Notion page, or a sentence about the flow.

    SHOP-388 · Delivery step
  2. 02

    Tokay designs the cases

    Every input split into classes and boundaries, written as Gherkin with example rows.

    10-digit mobile9 digitsempty
  3. 03

    Your team reviews

    Comment on any case. Tokay answers with a change you accept or reject.

    @Tokay split TC-02.3 into two rows
  4. 04

    Automate and approve

    One test per row, dry-run in a sandbox. A person approves the version.

    Approve version 13
Product

Built so a red result means your app broke

Coverage is computed, not claimed. Runs are isolated. Our own trouble never shows up as your failure.

Test design

Coverage you can see, per requirement

Each class once, every boundary, at most one invalid class per row, and pairs of valid classes where the risk is high.

AC1Delivery details are checked✓ Covered
+3 positive · −6 negative · 0 empty classes
AC2Card or cash on delivery✓ Covered
+5 positive · −3 negative · pairwise, high risk
AC3Promo code reduces the total✓ Covered
+4 positive · −4 negative · 0 empty classes
Review

Comment, and Tokay makes the change

@Tokay TC-02.3 has two faults at once. Split it.

− TC-02.3 · letters + empty+ TC-02.3 · letters+ TC-02.4 · 9 digitsAcceptReject
Runs

A failed run tells you why

agent.act add two products to the cart
app.open open '/cart'
agent.act pay with the saved cardWhy it failed
On a 375px screen the cart summary covers the Pay button.
Re-verify

Every Jira fix, checked three times

When a ticket moves to Ready for QA, VibeQA runs its repro suite three times and posts one verdict back. One lucky pass is not a fix.

JIRASHOP-412In progressReady for QA
Checkout button hidden on iPhone SE
Re-verify · 3 runs
  1. #48211m 38s
  2. #48221m 41s
  3. #48231m 36s
Fixed · 3 of 3 passed
Outcomes

Infra trouble is never your failure

  • Passed
  • Failed
  • Flaky
  • Infra error · retried
Isolation

A fresh sandbox for every run

  • Own browser and network
  • No platform secrets inside
  • Screenshots, trace and video for 7 days
Self-host

Your servers, your key

Docker Compose on one server, your own Anthropic key, apps that never leave your network.

$ docker compose up -d
Compare

What changes for your team

Test case spreadsheetHand-written PlaywrightVibeQA
Positive, negative and boundary cases per inputBy handBy handYes
Coverage gaps found for youNoNoYes
Reviewers suggest changes in commentsBy handNoYes
Runs on its own, each in a fresh sandboxNoBy handYes
Jira fixes re-checked three timesNoNoYes
Infrastructure errors kept apart from failuresNoNoYes
Pricing

Start free. Pay when you ship weekly.

AI usage is included in every Cloud plan.

Free

For trying it on one app

$0/month

Start free
  • 1 project
  • 100 runs a month
  • Test design and review with Tokay
  • Run history for 7 days
Most teams pick this

Team

For a team shipping every week

$49/month

Start a Team trial
  • 5 projects
  • 2,000 runs a month
  • Jira re-verify
  • Slack and Telegram notifications
  • CI API keys

Business

For several teams and products

Custom

Contact sales
  • Unlimited projects
  • More runs in parallel
  • Team-scoped projects and roles
  • Priority support
Self-hostFree. Docker Compose on your own server, your own Anthropic key, and apps that never leave your network.Read the deploy guide
FAQ

Questions teams ask first

Something else? Write to us

Can Tokay approve its own tests?

No. Tokay drafts the design, suggests changes and writes the code. A person approves every version, and approval needs complete coverage, no open comments and a finished dry run.

Where do the tests run, and what can they reach?

Each run gets a fresh sandbox with its own browser and network. It holds no platform secrets, only a token that expires with the run.

Which apps can VibeQA test?

Web apps today. Android and iOS are next.

Who pays for the AI?

On Cloud, AI usage is included in your plan. When you self-host, Tokay uses your own Anthropic key.

How does Jira re-verify work?

Connect Jira once. When a ticket with a repro suite moves to Ready for QA, VibeQA runs the suite three times and comments with one verdict: fixed, still reproducing, flaky or blocked.

What happens with flaky tests?

When some runs pass and others fail, the verdict is flaky and a person decides. VibeQA never reports a lucky pass as a fix.

Ship when Tokay says okay.

Tokay geckos are named for their call: to-kay. Ours makes it when your release is green.