Firecrawl Migration Deep Dive

Current State

!npm list puppeteer playwright cheerio 2>/dev/null | grep -E "puppeteer|playwright|cheerio" || echo 'No scraping libs found'

Overview

Migrate from custom scraping (Puppeteer, Playwright, Cheerio) or competing APIs to Firecrawl. Firecrawl eliminates browser management, anti-bot handling, and JS rendering infrastructure. This skill shows equivalent code for common scraping patterns.

Prerequisites

• An inventory of existing targets, permissions, extraction schemas, rate controls, and data-retention requirements.
• A staging environment with approved target pages or synthetic fixtures; do not use production credentials in migration tests.
• A rollback owner and a period in which the old path remains available for reconciliation.

Migration Comparison

Instructions

Step 1: Replace Puppeteer Single-Page Scrape

// BEFORE: Puppeteer (20+ lines, browser management)
import puppeteer from "puppeteer";

async function scrapePuppeteer(url: string) {
  const browser = await puppeteer.launch();
  const page = await browser.newPage();
  await page.goto(url, { waitUntil: "networkidle2" });
  const html = await page.content();
  const title = await page.title();
  await browser.close();
  return { html, title };
}

// AFTER: Firecrawl (5 lines, no browser needed)
import FirecrawlApp from "@mendable/firecrawl-js";

const firecrawl = new FirecrawlApp({ apiKey: process.env.FIRECRAWL_API_KEY! });

async function scrapeFirecrawl(url: string) {
  const result = await firecrawl.scrapeUrl(url, {
    formats: ["markdown"],
    onlyMainContent: true,
    waitFor: 2000,
  });
  return { markdown: result.markdown, title: result.metadata?.title };
}

Output

Keep a migration receipt listing each target, old and proposed behavior, validation result, permitted data destination, canary outcome, and rollback decision. Redact page content, credentials, and any personal data collected during comparison.

Examples

Choose one approved staging page and compare only metadata, content hash, and schema-valid field counts from the old and new paths. If the new path exceeds the credit budget, changes the extraction contract, or captures an unapproved field, disable the canary and retain the old path while the mapping is reviewed.

Step 2: Replace Cheerio HTML Parsing

// BEFORE: fetch + cheerio (manual parsing)
import * as cheerio from "cheerio";

async function scrapeCheerio(url: string) {
  const html = await fetch(url).then(r => r.text());
  const $ = cheerio.load(html);
  return {
    title: $("h1").first().text(),
    content: $("main").text(),
    links: $("a").map((_, el) => $(el).attr("href")).get(),
  };
}

// AFTER: Firecrawl with extract (LLM-powered, no CSS selectors)
async function extractFirecrawl(url: string) {
  const result = await firecrawl.scrapeUrl(url, {
    formats: ["extract", "links"],
    extract: {
      schema: {
        type: "object",
        properties: {
          title: { type: "string" },
          content: { type: "string" },
        },
      },
    },
  });
  return {
    title: result.extract?.title,
    content: result.extract?.content,
    links: result.links,
  };
}

Step 3: Replace Crawl Pipeline

// BEFORE: Playwright crawler (100+ lines, queue, browser pool)
// - launch browser pool
// - manage visited URLs set
// - extract links, enqueue
// - handle errors per page
// - close browsers on exit

// AFTER: Firecrawl crawl (10 lines)
async function crawlSite(baseUrl: string) {
  const result = await firecrawl.crawlUrl(baseUrl, {
    limit: 100,
    maxDepth: 3,
    includePaths: ["/docs/*", "/api/*"],
    excludePaths: ["/blog/*"],
    scrapeOptions: {
      formats: ["markdown"],
      onlyMainContent: true,
    },
  });

  return result.data?.map(page => ({
    url: page.metadata?.sourceURL,
    title: page.metadata?.title,
    content: page.markdown,
  }));
}

Step 4: Gradual Migration with Adapter Pattern

// Adapter interface for gradual migration
interface ScrapeAdapter {
  scrape(url: string): Promise<{ title: string; content: string }>;
  crawl(url: string, maxPages: number): Promise<Array<{ url: string; content: string }>>;
}

class FirecrawlAdapter implements ScrapeAdapter {
  private client: FirecrawlApp;

  constructor() {
    this.client = new FirecrawlApp({ apiKey: process.env.FIRECRAWL_API_KEY! });
  }

  async scrape(url: string) {
    const result = await this.client.scrapeUrl(url, {
      formats: ["markdown"],
      onlyMainContent: true,
    });
    return {
      title: result.metadata?.title || "",
      content: result.markdown || "",
    };
  }

  async crawl(url: string, maxPages: number) {
    const result = await this.client.crawlUrl(url, {
      limit: maxPages,
      scrapeOptions: { formats: ["markdown"], onlyMainContent: true },
    });
    return (result.data || []).map(page => ({
      url: page.metadata?.sourceURL || url,
      content: page.markdown || "",
    }));
  }
}

// Feature flag controlled migration
function getScrapeAdapter(): ScrapeAdapter {
  if (process.env.USE_FIRECRAWL === "true") {
    return new FirecrawlAdapter();
  }
  return new LegacyPuppeteerAdapter();
}

Step 5: Remove Old Dependencies

set -euo pipefail
# After migration is complete and verified
npm uninstall puppeteer puppeteer-core
npm uninstall playwright @playwright/test
npm uninstall cheerio

# Remove browser downloads
npx playwright uninstall --all 2>/dev/null || true

# Verify no lingering references
grep -r "puppeteer\|playwright\|cheerio" src/ --include="*.ts" || echo "Clean!"

Migration Checklist

• [ ] Install @mendable/firecrawl-js
• [ ] Create adapter layer wrapping Firecrawl
• [ ] Replace single-page scrapes with scrapeUrl
• [ ] Replace crawl loops with crawlUrl
• [ ] Replace HTML parsing with extract or markdown
• [ ] Feature flag to switch between old and new
• [ ] Run both in parallel, compare outputs
• [ ] Remove old scraping dependencies
• [ ] Delete browser management code

Error Handling

Resources

• Firecrawl vs Puppeteer
• Firecrawl Scrape Options
• Advanced Scraping Guide

Next Steps

For advanced troubleshooting, see firecrawl-advanced-troubleshooting.