Supercharged Testing: AI-Powered Workflows with Playwright + MCP - Debbie O'Brien

Debbie O’Brien, a senior staff developer relations engineer at Block and former Microsoft employee, presents a talk on supercharging end-to-end testing with AI, specifically using Playwright and the Model Context Protocol (MCP). She begins by introducing Playwright as an open-source testing library for automating web browser interactions, and then explains MCP as a protocol that gives LLMs the ability to interact with tools like browsers. The talk covers the evolution from using the Playwright MCP server for browser automation to the newer Playwright MCP test server, which is specifically designed for testing. She introduces Playwright Agents, which are specialized AI agents for planning, generating, and healing tests. The planner agent creates test plans by exploring the application, the generator writes tests based on those plans, and the healer fixes broken tests. She emphasizes the importance of seed files to guide the agents and avoid duplicated code. The talk includes live demos showing the agents in action, and discusses debugging tools like the HTML report, trace viewer, and UI mode, which now include a ‘copy prompt’ button to help AI debug failures. She also shares her personal experience of using AI to generate tests for a hotel booking website, highlighting common issues with AI-generated tests such as unnecessary timeouts and selector problems. The talk concludes with a vision of AI-driven testing workflows where humans are outside the loop, focusing on higher-level tasks.