browser automation
The agent drives a real Chrome — clicking, filling forms, reading console and network — so web QA runs on the actual rendered page.
What it is
Browser automation as MCP tools: the agent drives a real Chrome session — navigating, clicking, filling forms, reading console messages and network requests, capturing screenshots and recordings.
Why
QA through a real browser catches what a headless harness can't: the actual rendered page, the console errors, the network requests. If the agent can drive the same Chrome I use, its verification means something.
How I use it
Web QA and debugging run on the actual rendered page: reproduce the reported behavior, read the real console error, capture the broken state. Batched actions — navigate, settle, capture in one call — keep the loop fast.
Get it
Install the Claude in Chrome extension, or use any browser-automation MCP your client supports. Playwright covers the headless equivalent for scripted regression tests.
Connected
This thread is scoped to browser automation. Opening another node starts a new one.
Ask this node