Mimicry

AI-powered browser testing that learns what works, remembers verified runs, and adapts when your application changes.

AI-powered browser testing

Describe the test. Mimicry makes it repeatable.

Write browser tests in plain language or Gherkin-style steps. Mimicry uses AI to interpret the intent, find the right page elements, and generate executable Playwright interactions. After a successful run, it keeps a verified recording so later runs can repeat the behavior without another model call.

When an application change breaks a test, Mimicry can inspect the failure, reason about the updated interface, and repair the recorded steps. AI is used to learn or adapt; stable tests stay fast and deterministic.

View Mimicry on GitHub

A test in plain language

await mimicry`
  navigate to https://playwright.dev/
  click on "get started"
  and click on "trace viewer"
`;

How it works

Learn → Remember → Repeat → Troubleshoot & Fix

1. Learn

Interpret each instruction, reason about page structure, choose likely elements, and generate Playwright actions.

2. Remember

Save a verified recording of resolved interactions and the context needed to repeat them.

3. Repeat

Replay known-good behavior without AI calls while the application remains compatible.

4. Troubleshoot & Fix

When a run breaks, analyze the failure, relearn the changed interface, repair the steps, and save a new verified recording.

Technology

  • TypeScript
  • Node.js
  • Playwright
  • OpenAI-compatible providers

Includes semantic element selection, navigation and form interactions, test repair, and token-usage tracking.