Skip to content

Core Concepts

Core-concepts screenshot

Key objects

Application

The web application you want to test. For cloud execution, the application must be reachable over the public web. VibeTesting targets the DOM (standard HTML elements and structure); stacks that expose the DOM — for example React, Angular, Vue, or server-rendered pages — work as expected. UIs without DOM-backed controls (such as canvas-only interfaces) may not be testable. See Does my app need a DOM-based UI?.

Test run

One execution of a test in one of two modes:

  • Exploration (the agent decides safe next actions)
  • Guided (the agent follows your Tests Guidelines: flow goals or step-by-step)

Suite

A suite is a named collection of tests that you can re-run anytime. Suites help you organize testing by goal, for example:

  • Checkout process
  • Adding new items
  • Sanity before release

Suites make it easy to validate fixes and catch regressions by re-running the same flows after changes. You can trigger re-runs from the agent chat or via the External API.

Inventory

The Inventory gives you a central view of all suite executions, suites, and individual tests — with execution status, timestamps, and result metrics.

Execution modes

Exploration

In exploration mode:

  • You define focus areas and optional General Instructions
  • The agent samples safe actions to increase coverage
  • The agent reports UI and flow problems with severity and context

Guided

In guided mode:

  • You provide Tests Guidelines for each flow (either high-level goals or explicit step-by-step)
  • You can also provide General Instructions that apply to the whole run (constraints, known issues, focus areas)
  • The agent follows your intent strictly, staying on-path to complete the described journey
  • Best for testing specific user flows, whether you express them as goals or as step-by-step instructions

What the agent does (high level)

VibeTesting orchestrates multiple AI capabilities to:

  • Propose the next safe action (exploration)
  • Follow user-provided Tests Guidelines (guided)
  • Detect UI and functional issues (overlaps, broken flows, broken links, error states)
  • Locate the correct element to interact with on each step

Additional capabilities

  • Real-time logs — follow execution progress as it happens with streamed log entries
  • Batched execution — run tests in parallel batches to reduce total execution time
  • PRD upload — generate test flows automatically from your product requirements
  • Scheduling — run tests on a schedule: one-time, daily, or weekly, via chat or API
  • External API — trigger and manage executions from CI/CD or third-party systems