Labsco
MCP SERVER

CodingBaby Browser

by buyitsydney

Click a page by screen coordinate rather than by selector, and fold a whole form fill into one batched call.

Browser Automation, Control & ScreenshotsVerified
Summary
Aiming by coordinate makes the screenshot part of the loop, not an extra.

click takes a coordinate and type goes wherever focus already is, so the working rhythm is screenshot, aim, act, screenshot — and area_screenshot is there so each turn of that loop ships a rectangle instead of a full page. wait is built for the same rhythm: it returns the page state after the pause rather than only blocking for it. batch is the answer to what that loop costs in round trips, combining click, type and press_key into one command with an interval between operations.

What it is

A browser driver whose pointer tools take coordinates — 14 tools covering navigation, coordinate clicks, typing, area screenshots, viewport size and tabs.

What you get
  • Clicks placed at a coordinate, and text typed into whatever currently has focus
  • Screenshots of one rectangle of the page, given its top-left and bottom-right corners
  • A viewport set to a width and height you choose
  • A wait of a set number of seconds that returns a screenshot of the state it waited for
  • Several operations run in sequence in a single command, with an interval between them
  • Tabs opened, listed, selected by index and closed
Requirements

Nothing to supply — no account, no key. The pointer tools address the page by position through coordinate, topLeft and bottomRight, so aiming them means a screenshot and a viewport whose size you set; scroll is the one that takes a selector.

Setup effort

One command — npx @sydneyassistent/codingbaby-browser-mcp