Max Nardit
AI Agents

How to Send Web Pages to Cursor with Share2Agent

Share documentation, API references, and code examples directly to Cursor. Share2Agent extracts the page content, and a lightweight receiver saves it as markdown that you can reference in Cursor's AI chat with @file.

Prerequisites

  • Cursor IDE installed
  • Python 3.8+ (for the receiver script)
  • Share2Agent Chrome extension installed

How It Works

The approach is straightforward: a small Python receiver saves shared pages as markdown files in your project directory. Cursor can then read these files and use them as context in AI chat.

Chrome  -->  Share2Agent  -->  receiver.py  -->  .pages/article.md  -->  Cursor @file

Step 1: Set Up the Receiver

The receiver script from the Share2Agent repository works with any AI coding tool, not just Claude Code. Download it:

bash
curl -O https://raw.githubusercontent.com/mnardit/share2agent/main/examples/receiver/receiver.py
pip install pyyaml

Step 2: Configure the Save Directory

Run the receiver with PAGES_DIR pointing to a directory inside your Cursor project:

bash
PAGES_DIR=/path/to/your/project/.pages python3 receiver.py

Using .pages (with a leading dot) keeps shared content out of your source tree while remaining accessible to Cursor. Add it to .gitignore:

# .gitignore
.pages/

Step 3: Configure Share2Agent

  1. Click the Share2Agent extension icon in Chrome.
  2. Open Settings.
  3. Set the Webhook URL to http://localhost:9876.
  4. Save.

Step 4: Share a Page

  1. Navigate to a documentation page, API reference, or code example.
  2. Click the Share2Agent icon.
  3. Add an optional comment describing what you need (e.g., "how to use this hook").
  4. Click Share.

The receiver saves the page as a markdown file:

.pages/2026-03-28-1430-react-useeffect-docs.md

The file includes YAML frontmatter with the URL, title, author, and your comment, followed by the extracted content.

Step 5: Reference in Cursor

Open Cursor's AI chat (Cmd+L / Ctrl+L) and reference the saved file:

@.pages/2026-03-28-1430-react-useeffect-docs.md

How do I clean up effects in this pattern?

Cursor reads the full file content and uses it as context for the response.

Reference Multiple Files

If you have shared several pages for a research task, reference them all:

@.pages/2026-03-28-1430-react-useeffect-docs.md
@.pages/2026-03-28-1435-react-hooks-faq.md

Compare the approaches in these two docs.

Alternative: Use Cursor Rules

For pages you reference repeatedly, copy the content into a .cursor/rules file so it is automatically included as context:

bash
cp .pages/2026-03-28-1430-react-useeffect-docs.md .cursor/rules/react-effects.md

This way, Cursor always has that reference available without explicit @file mentions.

Alternative: MCP Integration

Cursor supports the Model Context Protocol (MCP). Instead of saving files, you could build a custom MCP server that:

  1. Receives webhooks from Share2Agent.
  2. Stores pages in memory.
  3. Exposes them as MCP resources that Cursor can query.

This is more advanced but avoids file clutter. See Cursor MCP docs for setup instructions.

Tips

  • Organize by project -- run separate receivers for different projects, each saving to its own .pages/ directory.
  • Use comments as instructions -- your Share2Agent comment is saved in the file's frontmatter. When you reference the file in Cursor, the AI sees your note alongside the content.
  • Clean up periodically -- delete old .pages/ files you no longer need to keep your project tidy.

What's Next?

  • Build a docs library -- share key documentation pages at the start of each project and reference them throughout development.
  • Share error pages -- when debugging, share the relevant Stack Overflow answer or GitHub issue directly and ask Cursor to apply the fix.
  • Combine with Composer -- reference shared pages in Cursor's Composer mode to generate code based on real documentation rather than training data.