← All Webinars

On-Demand

The New Playwright: Built for Agents

A hands-on tour of Playwright's agent-ready tooling: MCP vs. CLI myths busted, tests generated from a single prompt, agentic trace debugging, and agent-written tests turned into production monitoring.

Date

Jul

16

2026

Time

6 PM CEST

12 PM EST

9 AM PST

AI is writing code faster than it can be tested or monitored, and Playwright has become AI-native to keep up. In this recorded session, Stefan Judis shows why a coding agent without a real browser is just guessing, and what changes when you hand it one. Through live demos, he generates a complete end-to-end test from a single prompt, debugs a failing test straight from the trace file, and closes the loop by turning the agent-written test into production monitoring with Checkly.

What we'll cover

  1. Why an agent that can't open a browser is just guessing, and how page snapshots stop it.
  2. MCP vs. CLI, actually measured: a live token comparison that busts the "MCP eats your context window" myth.
  3. Generating a complete end-to-end test from a single prompt, then debugging it with npx playwright trace and an agent-attached debug session.
  4. Browser annotations for agent feedback, and captioned screencast videos as proof of work.
  5. From agent-written tests to production monitoring with Checkly, and Rocky AI's root-cause analysis on failures.

What you'll walk away with

  • —MCP and the CLI are just two interfaces to the same tool, token spend is no longer the deciding factor.
  • —An agent using a recent model and a real browser plans, generates, and heals. However, outdated skills and an old Playwright negatively impact its quality and speed.
  • —Checks are code, and agents write code: turn the tests your agent writes into production monitoring with Checkly and Rocky AI.

Speaker

Stefan Judis

Developer Relationships · Checkly