Scrape

Core Feature

Extract clean data from any single URL in multiple formats. Handles JavaScript, proxies, rate limits, and caching automatically.

Overview

The Scrape endpoint converts any web page into clean, LLM-ready output. One credit per page, with optional add-ons for JSON extraction, enhanced proxy, and more.

Formats Available

Format	Description	Extra Cost
markdown	Clean LLM-ready text	Included
html	Cleaned HTML	Included
rawHtml	Unmodified source HTML	Included
summary	AI-generated summary	Included
screenshot	Visual capture (full page, quality, viewport options)	Included
links	All page links	Included
images	All image URLs	Included
branding	Full brand identity/design system extraction	Included
json	Structured LLM extraction with schema	+4 credits
changeTracking	Detect changes vs previous scrape	See Change Tracking

Basic Usage

from firecrawl import Firecrawl

firecrawl = Firecrawl(api_key="fc-YOUR-API-KEY")

# Simple scrape — returns markdown by default
result = firecrawl.scrape("https://example.com")
print(result['markdown'])

import Firecrawl from '@mendable/firecrawl-js';

const firecrawl = new Firecrawl({ apiKey: "fc-YOUR-API-KEY" });
const doc = await firecrawl.scrape("https://example.com", {
  formats: ["markdown", "html"]
});
console.log(doc.markdown);

Multiple Formats

result = firecrawl.scrape(
    "https://example.com",
    formats=["markdown", "html", "screenshot", "links", "images"]
)
print(result['markdown'])
print(result['html'])
print(result['screenshot'])  # screenshot URL
print(result['links'])       # array of links
print(result['images'])      # array of image URLs

JSON Extraction (Schema-Based)

Extract structured data with a Pydantic model or JSON schema. In v2, the schema is embedded inside the format object:

from pydantic import BaseModel

class CompanyInfo(BaseModel):
    company_mission: str
    supports_sso: bool
    is_open_source: bool
    is_in_yc: bool

result = firecrawl.scrape(
    "https://firecrawl.dev",
    formats=[{"type": "json", "schema": CompanyInfo.model_json_schema()}],
    only_main_content=False,
    timeout=120000
)
print(result['json'])

v2 Format Change

In v2, use formats: [{"type": "json", "schema": {...}}]. The old v1 jsonOptions parameter no longer exists.

Prompt-Only Extraction (No Schema)

result = firecrawl.scrape(
    "https://firecrawl.dev",
    formats=[{"type": "json", "prompt": "Extract the company mission from the page."}],
    only_main_content=False
)

JSON Extraction Limitation

JSON extraction works on the markdown conversion — HTML attributes (data-id, custom attributes) are stripped and cannot be extracted. Workarounds:

• Use rawHtml format + parse client-side
• Use executeJavascript action to inject attribute values into visible text first

Branding Format

Extract a complete design system from any website:

result = firecrawl.scrape(
    url="https://firecrawl.dev",
    formats=["branding"]
)
print(result['branding'])

Returns: colorScheme, logo, colors (primary/secondary/accent/background/text), fonts, typography (fontFamilies/fontSizes/fontWeights/lineHeights), spacing, components (buttonPrimary/buttonSecondary/input), icons, images, animations, layout, personality.

Page Actions (Browser Interaction)

Interact with the page before scraping:

result = firecrawl.scrape(
    url="https://example.com/login",
    formats=["markdown"],
    actions=[
        {"type": "write", "text": "john@example.com"},
        {"type": "press", "key": "Tab"},
        {"type": "write", "text": "secret"},
        {"type": "click", "selector": 'button[type="submit"]'},
        {"type": "wait", "milliseconds": 1500},
        {"type": "screenshot", "full_page": True},
    ],
)

Action Types

Type	Description
wait	Wait for milliseconds
screenshot	Take screenshot (supports full_page)
scroll	Scroll the page
scrape	Scrape at current state
click	Click an element by CSS selector
write	Type text into focused element
press	Press a keyboard key
executeJavascript	Run custom JS on the page
generatePDF	Generate a PDF of the page

Caching (maxAge)

Default cache window is 2 days (172800000ms) — up to 5x faster scrapes:

# Use cache if less than 2 days old (default behavior)
result = firecrawl.scrape("https://example.com")

# Use cache if less than 10 minutes old
result = firecrawl.scrape("https://example.com", max_age=600000)

# Always get fresh content
result = firecrawl.scrape("https://example.com", max_age=0)

# Opt out of storing in cache
result = firecrawl.scrape("https://example.com", store_in_cache=False)

Location & Language

result = firecrawl.scrape(
    "https://example.com",
    location={"country": "DE", "languages": ["de"]}
)

Country defaults to US. Uses ISO 3166-1 alpha-2 codes.

Batch Scrape

Scrape multiple URLs in one operation:

job = firecrawl.batch_scrape(
    [
        "https://firecrawl.dev",
        "https://docs.firecrawl.dev",
    ],
    formats=["markdown"],
    poll_interval=2,
    wait_timeout=120
)

Async Batch

job = firecrawl.start_batch_scrape(["https://example.com", "https://example.org"])
status = firecrawl.get_batch_scrape_status(job.id)

Batch scrape jobs expire after 24 hours.

Batch Endpoints

Method	Path	Description
POST	/v2/batch/scrape	Start batch scrape
GET	/v2/batch/scrape/{id}	Get batch status
DELETE	/v2/batch/scrape/{id}	Cancel batch scrape
GET	/v2/batch/scrape/{id}/errors	Get batch errors

Cost

Operation	Credits
Basic scrape	1 per page
JSON mode	+4 per page
Enhanced proxy	+4 per page
PDF parsing	+1 per PDF page
Screenshot	Included
Branding	Included

---

Next: Search →