Files
2026-01-21 15:49:34 +08:00
..
2026-01-21 15:49:34 +08:00

Generate Scraper Script with Claude Code + Actionbook

This demo shows how Actionbook helps AI Agents (like Claude Code) generate more accurate web scraping scripts by providing verified selectors and page structure information.

Demo

Watch how Actionbook helps Claude Code write a correct scraper, 10x faster.

https://github.com/user-attachments/assets/912b7d39-1e55-43b7-b766-344b242762e9

The Problem

When you ask Claude Code to "write a scraper for X website", it needs to figure out:

  • What CSS selectors to use
  • How the page structure works
  • How to handle dynamic content
  • Which elements contain the data you need

Without Actionbook, Claude Code has to guess selectors based on common patterns, often resulting in:

  • Wrong or fragile selectors
  • Missing data fields
  • Scripts that break when the page structure changes
  • Multiple rounds of trial and error

With Actionbook, Claude Code gets verified selectors and page structure directly, generating working scripts on the first try.

Quick Start

1. Install Actionbook

Install via plugin

claude plugin marketplace add actionbook/actionbook
claude plugin install actionbook@actionbook-marketplace

Install manually

claude mcp add actionbook -- npx -y @actionbookdev/mcp@latest
npx skills add actionbook/actionbook

2. Add to Your Prompt

When asking Claude Code to write a scraper, add this to your prompt:

Use Actionbook to understand the page before taking action.

That's it! Claude Code will automatically query Actionbook for verified selectors and page structure.

Comparison

Without Actionbook

See comparasion/cc-without-actionbook/

  • Claude Code guessed selectors like .company-list-card-medium-large
  • Had to visit each company's detail page individually
  • Took multiple attempts to find working selectors
  • Result: Only 25 companies scraped, some with incomplete data

With Actionbook

See comparasion/cc-with-actionbook/

  • Claude Code got verified selectors from Actionbook
  • Knew exactly how to expand cards and extract details
  • Generated working script on first attempt
  • Result: All 194 companies scraped with complete data