Labsco
MCP SERVER

YetiBrowser MCP

by yetidevworks

Drive a page through its accessibility tree: snapshot it, click, hover, type or pick from a dropdown by element reference, then read the console back.

Browser Automation, Control & ScreenshotsVerified
Summary
Element references come out of a snapshot, so every interaction starts with one.

The interaction tools take an element and a ref rather than a CSS selector, and those refs come from browser_snapshot — which means a click after the page changes needs a fresh snapshot first. What comes back is text-first: the snapshot and the console logs carry the state, and browser_screenshot is there for the cases a picture settles faster than a tree. Waiting is expressed as a number of seconds rather than a condition, so a slow page is handled by choosing a duration.

What it is

A browser-driving server built around accessibility snapshots — 12 tools for navigation, input, waiting and readback.

What you get
  • Move around: open a URL, go back, go forward.
  • Take an accessibility snapshot of the page, which is what produces the element references every interaction tool expects.
  • Act on a referenced element — click it, hover it, type into it with an optional submit, or choose values in a dropdown.
  • Press a named key, and wait a set number of seconds when the page needs the time.
  • Read the result back as the browser console logs, or as a screenshot of the current page.
Requirements

A browser for the server to drive. No account, no key.

Setup effort

One command — npx -y @yetidevworks/server