@0xkobold/pi-cloudflare-browser

🌐 Pi Cloudflare Browser

Cloudflare Browser Rendering for pi-coding-agent.

Web scraping, screenshots, and PDF generation via Cloudflare's Browser Rendering API.

Features

• 🕷️ Web crawling - Multi-page with depth control
• 📸 Screenshots - Inline display + saved files
• 📄 PDF generation - From any URL
• 📄 Content extraction - Rendered HTML/Markdown
• 🎯 Standalone - Works without 0xKobold
• 🔌 Integrable - Exposes client for custom use

Installation

# Via pi CLI
pi install npm:@0xkobold/pi-cloudflare-browser

# Or in pi-config.ts
extensions: ['npm:@0xkobold/pi-cloudflare-browser']

Setup

1. Get Cloudflare credentials:

   • Account ID: https://dash.cloudflare.com (right sidebar)
   • API Token: My Profile → API Tokens → Create → "Browser Rendering" template

2. Configure:

export CLOUDFLARE_API_TOKEN=your_token
export CLOUDFLARE_ACCOUNT_ID=your_account_id
export PI_OUTPUT_DIR=./outputs  # Optional: where to save files

Usage

Standalone

import { CloudflareClient, loadConfig } from '@0xkobold/pi-cloudflare-browser';

// Load from env
const config = await loadConfig();
const client = new CloudflareClient(config);

// Take screenshot
const result = await client.screenshot('https://example.com', { fullPage: true });
// result.screenshot = base64 image
// result.metadata = { title, url, status }

// Crawl website
const crawlId = await client.crawl({
  url: 'https://example.com',
  limit: 10,
  render: true,
});

// Get crawl results
const crawl = await client.getCrawlResult(crawlId);

With pi-coding-agent

// Extension auto-registers tools:
// /cf_crawl      - Multi-page crawling
// /cf_screenshot - Capture screenshot
// /cf_pdf        - Generate PDF
// /cf_fetch      - Single page content

Tools

API

CloudflareClient

class CloudflareClient {
  async crawl(params: CrawlParams): Promise<string>;      // Returns job ID
  async getCrawlResult(jobId: string): Promise<CrawlResult>;
  async screenshot(url: string, opts?: ScreenshotOptions): Promise<ScreenshotResult>;
  async pdf(url: string, opts?: PdfOptions): Promise<PdfResult>;
  async fetchContent(url: string, opts?: ContentOptions): Promise<ContentResult>;
}

Configuration

Standalone Example

// screenshot.ts - No pi-coding-agent needed
import { CloudflareClient, loadConfig } from '@0xkobold/pi-cloudflare-browser';

async function main() {
  const config = await loadConfig();
  if (!config.enabled) {
    console.log('Set CLOUDFLARE_API_TOKEN and CLOUDFLARE_ACCOUNT_ID');
    return;
  }

  const client = new CloudflareClient(config);
  
  const result = await client.screenshot('https://example.com', {
    fullPage: true,
  });

  // Save to file
  await Bun.write('screenshot.png', Buffer.from(result.screenshot, 'base64'));
  console.log('Saved!');
}

main();

Links

• GitHub
• Cloudflare Browser Rendering Docs

License

MIT