← Back to Discover

onUI

Open source browser extension and local MCP server for marking up web pages with annotations or drawings and passing that structured context to AI agents. Works in Chrome, Edge and Firefox.

MCP serversUI design
Visit website ↗

What it does

onUI lets you annotate or draw on any web page and export what you marked as structured context for an AI agent. You pick elements or sketch regions, add an intent and a severity to each note, and copy the result out, with no changes to the app's own code. It has two parts: a browser extension for Chrome, Edge and Firefox, and a local MCP server with a native bridge, so agents can read the annotations without a cloud backend. The installer registers the server automatically for Claude Code and Codex when those CLIs are installed, and a manual JSON entry works for other MCP clients. Annotation is switched on per tab and is off by default. Output comes in four levels of detail, from compact to forensic. The extension installs from the Chrome Web Store or Edge Add-ons, or from a release installer, which is also the route for Firefox.

Capabilities

  • Annotate selected elements with intent and severity, or draw rectangles and ellipses over regions
  • Shift-click to select several elements and annotate them together
  • Export in compact, standard, detailed or forensic formats
  • Local MCP server with no cloud backend required
  • Per-tab on and off control, off by default

When to use it

  • You want to point an AI agent at specific UI problems on a live page
  • You use Claude Code, Codex or another MCP client and want to hand it visual feedback
  • You want to review a front end without changing its code