Browser Automation

import { Aside } from ‘@astrojs/starlight/components’;

Ptah’s browser automation suite lets agents drive a real Chrome or Chromium instance on your machine. It is built on top of Playwright and exposed through a dedicated set of MCP tools, so any agent — local or remote — can use it through the same uniform interface.

What you can do

Launch a headed or headless Chromium instance under Ptah’s control
Navigate to URLs and wait for load states deterministically
Click, type, scroll, and evaluate arbitrary JavaScript in the page
Capture full-page or viewport screenshots in PNG, JPEG, or WebP
Monitor every HTTP request and response the page makes
Record an animated GIF of a session for bug reports or demos

Tool catalog

Tool	Purpose
`ptah_browser_status`	Report whether a browser instance is currently running
`ptah_browser_navigate`	Open a URL and wait for the page to settle
`ptah_browser_click`	Click a selector or coordinate
`ptah_browser_type`	Type text into an input
`ptah_browser_content`	Return the rendered HTML or a readable text extract
`ptah_browser_evaluate`	Run JavaScript in the page context
`ptah_browser_screenshot`	Capture the current viewport or full page
`ptah_browser_network`	List captured HTTP requests and responses
`ptah_browser_record_start`	Begin recording an animated GIF
`ptah_browser_record_stop`	Stop recording and save the GIF
`ptah_browser_close`	Shut the browser down cleanly

Typical workflow

Agent calls ptah_browser_navigate with a target URL.
Agent inspects the DOM with ptah_browser_content or a targeted ptah_browser_evaluate.
Agent interacts with ptah_browser_click and ptah_browser_type.
Agent captures a ptah_browser_screenshot or reviews network traffic via ptah_browser_network.
Agent closes the browser with ptah_browser_close when the task is done.

See the dedicated pages in this section for full details of each operation.