Context Cue

Visual annotation tool for AI coding agents. Context Cue lets you annotate elements on your page with feedback, bug reports, feature requests, or any context. Annotations are automatically sent to Cursor via MCP, giving your AI agent the exact context it needs to understand and act on your feedback.

How it works

https://example.com/dashboard

Dashboard

Total Users
12,345
Revenue
$45,678
  1. Press Cmd+Shift (or Ctrl+Shift) to activate annotation mode
  2. Click any element on the page and enter your feedback
  3. Annotation is automatically sent to Cursor via MCP as context

What gets captured

  • Element selector — Class names, IDs, and DOM paths like .sidebar > button.primary
  • Screenshot — Visual context of the annotation
  • Console errors — JavaScript errors that occurred (if any)
  • Network failures — Failed API requests (if any)
  • Device info — Viewport size, user agent, browser, etc.
  • Your description — Any feedback, bug report, feature request, or context

Use cases

  • Bug reports — Report issues with specific elements
  • Feature requests — Suggest improvements or new features
  • Design feedback — Provide UI/UX feedback on specific components
  • General context — Give your AI agent context about any element

MCP integration

Context Cue integrates with Cursor via the Model Context Protocol (MCP). When you create an annotation, it's automatically available to your Cursor agent as context. The agent can use the selectors, screenshots, and your description to understand and act on your feedback.

See the Installation page for setup instructions.

Context Cue is an adaptation of Agentation, a dev tool by Benji Taylor, Dennis Jin, and Alex Vanderzon, with an MCP server integration.