Using MCP agent to observe Playwright debug browser session

[mrrinatino]

I have an idea for how I could use your MCP server.
I’d like the agent to observe the browser that is opened in debug mode when running Playwright tests. For example, when I run npx playwright test --debug, the browser launches and I step through the scenario interactively.

I want the agent to analyze each step — DOM tree, console output, network activity — and then adjust or improve the generated test code accordingly (e.g., suggest a better locator or detect errors).

This would be part of an automated workflow where an agent helps create an end-to-end test from scratch.

Is it possible to make the agent and Playwright use or observe the same browser instance/session for this purpose?