Skip to main content

​
Natural language + cache function

Playmatic tests use a combination of natural language with Playwright-compatible cache functions. Natural language instructions are used by a computer use agent to self-heal tests and complete actions that are difficult to test with stable selectors (e.g., smart waiting, smart assertions, or dynamic UIs). Cache functions make test actions fast and deterministic. 
// natural language step as a fall back action
testStep('Create first label', async ({ page }) => {

    // playwright cache function
    const createLabelButton = page.getByRole('button', { name: /create.*label/i }).or(
      page.getByRole('button', { name: /new label/i }).or(
        page.getByRole('button', { name: /add.*label/i })
      )
    );
    await createLabelButton.click();
  }, {
    // toggle to turn off self-healing when cache exists
    cacheOnly: true
  });
Combined, this framework gives developers the control to write fast tests that leverage the reasoning capabilities of agents.

​
Self-healing

The cache function is always executed first. If the cache function fails, a computer use agent will take action using the natural language instruction. If you prefer the test to not self-heal, this is a parameter that can be configured. Learn more in the writing tests guide.

​
Creating Tests

We recommend using a coding agent (like Cursor or Claude Code) with our automatically loaded context file to create tests. The context file provides your coding agent with complete SDK documentation, testing patterns, and best practices to generate robust Playmatic tests. You can always further refine the generated tests using custom code and Playmatic primitives.

​
Next steps

Using Claude Code

Learn how to write tests using Claude Code with our context file.

Writing a test

Learn more about the core concepts to writing a Playmatic test.