Releaseworthy

How it works

Agents do the volume. A person makes the call.

The agents run on Claude and are built with the Claude Agent SDK. A QA engineer reviews everything they produce and decides Go or No-Go.

Diagram of the workflow

Claude agents

  1. Plan Read the product, specs and changes, and draft a risk-based test plan.
  2. Write Write tests in your repository and open them as pull requests.
  3. Run Run the suites in your CI and collect evidence for every failure.
  4. HealUpdate tests that broke because the product changed, and flag failures that look like real bugs.

The loop repeats with every change.

QA engineer

  1. Reviews every change Nothing the agents write is merged unchecked.
  2. Owns the release call Confirms bugs, weighs the risks and decides Go or No-Go.

GONO-GOA signed report before each release.

How the work flows: every change the agents make passes a person before it reaches your main branch.

Onboarding

Getting started

  1. Step 1

    An intro by email

    You describe the product, the platforms, how you release and what worries you. A call follows if it helps.

  2. Step 2

    A short written agreement

    Scope and notice terms are agreed in writing before any work starts.

  3. Step 3

    Access, and only what is needed

    Usually: the right to open pull requests in your repository, access to CI, a test environment and test accounts. Secrets stay in your vault or CI secrets.

  4. Step 4

    A risk-based test plan

    We list your user flows, rank them by how much a failure would hurt, and agree what gets covered first.

  5. Step 5

    The first tests land

    The agents write the first tests as pull requests. The QA engineer reviews each one before it is merged.

Weekly rhythm

What a normal week looks like

Work happens async in your team's chat (Slack, Discord, Telegram or similar), in GMT+7 working hours with overlap for US and EU teams.

  • Every pull request

    The suites run in your CI. A red result is looked at by a person before anyone is asked to act on it.

  • New features

    Exploratory testing by hand, then new automated tests for the flows that matter.

  • Failures

    A real bug gets a report with steps and evidence. A test broken by a product change gets fixed. A flaky test gets traced to its cause.

  • Every week

    A written digest in your team's chat: what changed in coverage, open bugs and the risks we see.

Release gate

A Go/No-Go report before each release

The report says what was tested, what passed and failed, which bugs are open and how severe they are, what risk is left and what we recommend.

The QA engineer owns the call and signs the report. A No-Go names exactly what blocks the release and what would change the call. Shipping stays your decision; the report makes the risk visible before you make it.

What a No-Go looks like

NO-GO
  • Blockerpayment fails for saved cards
  • Evidencesteps, video, logs
  • Changes the callfix + retest

Example only, not client data.

Who does what

The split between agents and a person

Claude agents

  • Draft test plans from your specs, tickets and code
  • Write new tests and update tests the product has outgrown
  • Run the suites and collect evidence for each failure
  • Propose fixes for broken tests

The QA engineer

  • Sets the strategy and the priorities with you
  • Reviews every change before it is merged
  • Confirms each bug and its severity
  • Owns the Go/No-Go call before each release

Tools

What the work is built with

The agents are built with the Claude Agent SDK and run on Claude. The tests they write use standard, open tools, so your team can read and run them without us.

  • Web

    Playwright + TypeScript

    End-to-end tests in Playwright + TypeScript. Playwright Test runs tests in parallel across Chromium, Firefox and WebKit.

  • API

    Endpoints and journeys

    Tests for single endpoints and multi-step journeys, written in TypeScript and run in the same pipeline as the web tests.

  • Mobile

    Appium, Maestro or Mobilewright

    iOS and Android tests. We choose the framework that fits your app best:

    • AppiumOpen-source UI automation for iOS, Android and other platforms. Appium 3 was released in August 2025.
    • MaestroOpen-source mobile UI testing, with flows written as human-readable YAML.
    • MobilewrightTypeScript tests for iOS and Android with a Playwright-style API, from the makers of Mobile MCP.

Everything runs in your own repository and CI, with your issue tracker for bugs. The agent framework itself is open source: see agentic-qa.

Tell us what you are shipping next

One email is enough to start: your product, your platforms and how often you release. You get a plain answer on fit and price.

Email usSee pricing

Or write to qa@releaseworthy.app.