TutorialFeb 22, 20262 min read

Getting Started with AI-Generated Tests

Lastest's agentic AI crawls your application, interacts with elements, and generates visual test suites automatically while a human oversight dashboard lets you review and approve each test. Your first run consumes AI tokens to build the test suite, but every replay afterward is completely free.

Two-column comparison: the first run spends AI tokens to crawl, write Playwright, and capture baselines once per suite; every replay afterward is plain Playwright costing $0, runnable thousands of times a day. — You pay AI tokens once to build the suite; every replay afterward is zero-token plain Playwright at $0.

How it works

When you start an AI-generated test run, Lastest's agent navigates your application like a real user. It clicks buttons, fills forms, opens menus, and captures screenshots at each meaningful state change.

The agent uses a combination of DOM analysis and visual understanding to identify interactive elements and determine which user flows are worth testing. It prioritizes paths that cover the most visual surface area.

Left-to-right pipeline: the agent crawls the app like a user, captures screenshots, a human reviews the suite, and it becomes an approved baseline; AI runs only once at suite-build time so every CI replay afterward is free. — The agent crawls, captures, and proposes a suite; one human review turns it into an approved baseline, and AI only runs at this build stage.

Step 1: Configure your target

Point the agent at your application's URL and specify any authentication credentials it needs. You can also provide a sitemap or a list of entry points to guide the crawl.

lastest ai-generate \
  --url https://staging.yourapp.com \
  --auth-cookie "session=abc123" \
  --max-depth 3

Step 2: Review the generated suite

The human oversight dashboard shows every test the agent generated, including the steps it took and the screenshots it captured. You can approve, reject, or modify individual tests before they become part of your baseline.

Timeline of three milestones: 1) configure target URL and auth, 2) review the generated suite and approve or reject, 3) run on CI on every PR for $0. — Getting started is three steps: configure the target, review and approve the generated suite, then run it free on CI.

Step 3: Run on CI

Once approved, the test suite runs on every pull request - no AI tokens needed. It replays the exact same interactions and compares screenshots against your approved baselines.

If your UI changes intentionally, update the baselines with a single click. If something breaks, you'll see the diff immediately.