Imported from jeremylongshore/tons-of-skills-marketplace (skills/.curated/firecrawl-cost-tuning/SKILL.md). Install upstream with npx skills add jeremylongshore/tons-of-skills-marketplace --skill firecrawl-cost-tuning. Copyright stays with the author (MIT).

Firecrawl Cost Tuning

Overview

Firecrawl charges credits per operation: 1 credit per scrape, 1 per crawled page, 1 per map call, and variable credits for extract (LLM usage). An unbounded crawl on a large site can consume thousands of credits in minutes. This skill covers concrete techniques to reduce credit consumption by 50-80%.

Prerequisites

• Current provider pricing and contract terms, reviewed by the billing owner rather than assumed from examples.
• An approved domain allowlist, crawl budget, usage owner, and aggregate metering source.
• Synthetic or staging targets for validating limits before changing a production job.

Output

Record a cost-control receipt with the measurement window, approved target set, configured limits, aggregate credits consumed, owner decision, and follow-up date. Do not include credentials, page contents, or customer data.

Examples

Run a canary crawl against an approved staging target with a low page limit. Compare only aggregate credit use and valid-page count, then stop promotion if the observed cost or target expansion exceeds the approved budget.

Credit Cost Table

Instructions

Step 1: Always Set Crawl Limits

import FirecrawlApp from "@mendable/firecrawl-js";

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

// BAD: no limit — could crawl 100K pages
await firecrawl.crawlUrl("https://docs.large-project.org");
// Cost: potentially 100,000+ credits

// GOOD: bounded crawl
await firecrawl.crawlUrl("https://docs.large-project.org", {
  limit: 50,               // max 50 pages
  maxDepth: 2,             // only 2 levels deep
  includePaths: ["/api/*"], // only API docs
  excludePaths: ["/blog/*", "/changelog/*"],
  scrapeOptions: { formats: ["markdown"] },
});
// Cost: max 50 credits

Step 2: Use Scrape for Known URLs Instead of Crawl

// If you know which pages you need, don't crawl — scrape them directly
const targetUrls = [
  "https://docs.example.com/api/auth",
  "https://docs.example.com/api/users",
  "https://docs.example.com/api/billing",
];

// Cost: 3 credits (one per page)
const results = await firecrawl.batchScrapeUrls(targetUrls, {
  formats: ["markdown"],
});

// vs crawling the whole docs site: potentially 500+ credits

Step 3: Map First, Then Selective Scrape

// Map costs 1 credit and returns up to 30K URLs
const map = await firecrawl.mapUrl("https://docs.example.com");
// Cost: 1 credit

// Filter to only what you need
const apiDocs = (map.links || []).filter(url => url.includes("/api/"));
console.log(`${map.links?.length} total URLs, only ${apiDocs.length} are API docs`);

// Scrape only relevant pages
const results = await firecrawl.batchScrapeUrls(apiDocs.slice(0, 20), {
  formats: ["markdown"],
});
// Cost: 1 (map) + 20 (scrape) = 21 credits
// vs blind crawl: could be 500+ credits

Step 4: Cache to Prevent Re-Scraping

import { createHash } from "crypto";

const cache = new Map<string, { content: string; timestamp: number }>();
const CACHE_TTL = 24 * 3600 * 1000; // 24 hours

async function cachedScrape(url: string): Promise<string> {
  const key = createHash("md5").update(url).digest("hex");
  const cached = cache.get(key);

  if (cached && Date.now() - cached.timestamp < CACHE_TTL) {
    return cached.content; // Free — no API call
  }

  const result = await firecrawl.scrapeUrl(url, { formats: ["markdown"] });
  if (result.markdown) {
    cache.set(key, { content: result.markdown, timestamp: Date.now() });
  }
  return result.markdown || "";
}
// Typical savings: 50-80% credit reduction for recurring scrapes

Step 5: Monitor Credit Consumption

set -euo pipefail
# Check current credit balance
curl -s https://api.firecrawl.dev/v1/team/credits \
  -H "Authorization: Bearer $FIRECRAWL_API_KEY" | jq .

// Daily credit tracker
class CreditBudget {
  private dailyLimit: number;
  private usage = new Map<string, number>();

  constructor(dailyLimit = 1000) {
    this.dailyLimit = dailyLimit;
  }

  canAfford(estimatedCredits: number): boolean {
    const today = new Date().toISOString().split("T")[0];
    const used = this.usage.get(today) || 0;
    return used + estimatedCredits <= this.dailyLimit;
  }

  record(credits: number) {
    const today = new Date().toISOString().split("T")[0];
    this.usage.set(today, (this.usage.get(today) || 0) + credits);
  }

  remaining(): number {
    const today = new Date().toISOString().split("T")[0];
    return this.dailyLimit - (this.usage.get(today) || 0);
  }
}

const budget = new CreditBudget(1000);

// Before each crawl
if (!budget.canAfford(50)) {
  throw new Error(`Daily credit budget exceeded. ${budget.remaining()} credits left`);
}
await firecrawl.crawlUrl(url, { limit: 50 });
budget.record(50);

Step 6: Choose Minimal Formats

set -euo pipefail
# Cheapest: markdown only (1 credit, fastest)
curl -X POST https://api.firecrawl.dev/v1/scrape \
  -H "Authorization: Bearer $FIRECRAWL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com","formats":["markdown"]}'

# Avoid requesting screenshots, rawHtml, or extract unless needed
# Extract uses LLM calls — significantly more credits

Error Handling

Cost Optimization Summary

Resources

• Firecrawl Pricing
• Firecrawl Dashboard
• Rate Limits

Next Steps

For reference architecture, see firecrawl-reference-architecture.