Firecrawl Lean

Firecrawl CLI — lean reference

Check status: firecrawl --status (shows auth + credits + concurrency limit). Not installed? Run: npm install -g firecrawl-cli && firecrawl auth Auth commands: firecrawl login, firecrawl logout, firecrawl view-config

WebFetch vs Firecrawl

Try WebFetch first (free). Use Firecrawl when:

• Need web search (WebFetch can't search)
• JS-heavy SPA / marketing pages
• Full article content (WebFetch truncates long pages)
• WebFetch output < 1000 chars or looks incomplete

Escalation pattern

search → scrape → map+scrape → crawl* → interact*

* = advanced, install dedicated skill if needed

Commands

Search — no URL yet, find pages:

firecrawl search "query" --limit 5 -o .firecrawl/results.json --json
firecrawl search "query" --scrape -o .firecrawl/results.json --json        # include full content
firecrawl search "query" --sources news --tbs qdr:d --json                 # news, past day
firecrawl search "query" --country US --location "New York" --json         # geo-targeted

Scrape — have a URL:

firecrawl scrape "<url>" -o .firecrawl/page.md
firecrawl scrape "<url>" --only-main-content -o .firecrawl/page.md        # strip nav/footer
firecrawl scrape "<url>" --wait-for 3000 -o .firecrawl/page.md            # wait for JS
firecrawl scrape "<url>" --query "What is the enterprise price?"            # targeted Q (5 credits)
firecrawl scrape "<url>" --format markdown,html,screenshot -o .firecrawl/page.md  # multi-format
firecrawl scrape "<url>" --format json --schema '{"price":"string"}' -o .firecrawl/data.json  # structured

Map — find URLs within a site:

firecrawl map "<url>" --search "authentication" -o .firecrawl/urls.txt
firecrawl map "<url>" --limit 200 --json -o .firecrawl/urls.json
firecrawl map "<url>" --include-subdomains --sitemap -o .firecrawl/urls.json

Crawl — bulk extract site section (advanced — install firecrawl-crawl for full docs):

firecrawl crawl "<url>" --include-paths /docs --limit 50 --wait -o .firecrawl/crawl.json
firecrawl crawl "<url>" --max-depth 3 --delay 500 --exclude-paths /blog --wait -o .firecrawl/crawl.json

Download — save site as local files (advanced):

firecrawl download "<url>" --limit 20 -y -o .firecrawl/site/

Interact — click/fill forms after scrape (advanced — install firecrawl-instruct for full docs):

firecrawl scrape "<url>"
firecrawl interact --prompt "Click the login button"
firecrawl interact stop

Agent — AI-powered structured extraction (advanced — install firecrawl-agent for full docs):

firecrawl agent "extract all pricing tiers" --wait -o .firecrawl/data.json
firecrawl agent "extract pricing" --urls "<url>" --schema '{"tiers":[]}' --wait -o .firecrawl/data.json

Browser — CDP/Playwright-style automation (distinct from interact):

firecrawl browser launch-session "<url>"
firecrawl browser execute --code "document.querySelector('button').click()"
firecrawl browser list
firecrawl browser close <session-id>

Output rules

• Always write to .firecrawl/ with -o to avoid context bloat
• Add .firecrawl/ to .gitignore
• Always quote URLs (? and & are shell special chars)
• Never read entire output at once: head -50 file.md or grep "keyword" file.md
• Naming: .firecrawl/search-{query}.json, .firecrawl/{site}-{path}.md

Extract URLs from JSON results

jq -r '.data.web[].url' .firecrawl/search.json

Credits

firecrawl credit-usage

<p align="center"> <strong>firecrawl-lean</strong> </p>

<p align="center"> <img src="./docs/assets/firecrawl-lean-mark.svg" alt="firecrawl-lean mark" width="120" height="120"> </p>

<h1 align="center">Token-efficient Firecrawl skill for Coding Agents.</h1>

<p align="center"> Web search, scrape, crawl, and extract — ~60% fewer tokens than the official skill. </p>

<p align="center"> <img alt="claude-code" src="https://img.shields.io/badge/claude--code-skill-111111"> <img alt="token-efficient" src="https://img.shields.io/badge/tokens-~60%25_fewer-ff4d00"> <img alt="firecrawl" src="https://img.shields.io/badge/powered_by-firecrawl-222222"> <img alt="license" src="https://img.shields.io/badge/license-MIT-444444"> </p>

---

What this is

A single-file Claude Code skill that gives Claude the full Firecrawl CLI — web search, scraping, crawling, structured extraction, browser automation — in a form optimized for token efficiency.

The official Firecrawl skill set ships as 5 separate skills (firecrawl, firecrawl-crawl, firecrawl-agent, firecrawl-instruct, firecrawl-browser). Each loads independently, repeats shared context, and costs tokens every time Claude reads them. This skill merges all five into one lean reference.

Why lean matters

Every skill file loads into Claude's context window on each turn. Bloated skills = wasted tokens = higher cost + slower responses.

Fewer tokens in context = more room for your actual task.

Install

npx skills add https://github.com/alexsmedile/firecrawl-lean

Requires Firecrawl CLI:

npm install -g firecrawl-cli && firecrawl login

What's covered

Escalation pattern

search → scrape → map + scrape → crawl → interact / browser / agent

Start cheap. Escalate only when needed. The skill tells Claude when and why to step up.

Token efficiency techniques

• One file, no repetition — shared context (output rules, auth, credits) written once
• Examples over prose — commands show usage; no paragraphs explaining what --limit does
• Advanced commands stubbed — crawl/interact/agent/browser are present but not padded; install dedicated skills only if you need them
• Output rules baked in — Claude always writes to .firecrawl/ and never dumps raw output to context

Comparison with official skills

The official Firecrawl skills are excellent and full-featured. This skill is for users who:

• Run Claude Code frequently and care about context window cost
• Want a single install, not five
• Don't need the extended documentation — just the commands

If you need deep crawl configuration, browser profiles, or parallel agent jobs, the official skills have more detail.

Credits

Built on top of Firecrawl and the Firecrawl CLI. This skill is an independent, community-maintained efficiency wrapper — not affiliated with Firecrawl.