FIRE-1 Agent Model

Updated Feb 2026

FIRE-1 is an AI agent that enhances Firecrawl's extraction capabilities through intelligent browser automation. It plans and executes multi-step actions to uncover data that standard scraping cannot reach.

Overview

FIRE-1 differs from the Spark-1 models used in the Agent API. While Spark-1 focuses on autonomous web research (searching the web, navigating sites, synthesizing information), FIRE-1 is a browser automation agent designed to interact with complex page elements to extract structured data from specific URLs.

FIRE-1 vs Spark-1

Capability	FIRE-1	Spark-1 Mini/Pro
Purpose	Browser automation extraction	Autonomous web research
URLs Required	Yes	No (optional)
Endpoint	/v1/extract	/v2/agent
Browser Control	Full (clicks, forms, pagination)	Search + navigate
Schema Output	Yes	Yes
Web Search	No	Yes (built-in)
Cost	Variable (per-request complexity)	Dynamic (credit-based)
Best For	Complex single-site extraction	Multi-site research

When to Use FIRE-1

Use FIRE-1 when your target data requires browser interaction to access:

• Paginated content -- tables, forums, or listings that span multiple pages
• Dynamic elements -- data behind "Load More" buttons, accordions, tabs
• Multi-step navigation -- content requiring sequential clicks to reach
• Interactive forms -- data that appears after filter/search interactions
• Hidden content -- elements that only render after user interaction

Use Spark-1 (Agent API) instead when:

• You do not have specific URLs
• You need web search capabilities
• The task involves cross-site research
• Data is on simple, static pages

API Endpoint

POST https://api.firecrawl.dev/v1/extract

v1 Endpoint

FIRE-1 uses the /v1/extract endpoint, not /v2/extract. This is distinct from the newer Extract API which uses /v1/extract as well but can operate without an agent.

Rate Limit: 10 requests per minute

Parameters

Parameter	Type	Required	Description
urls	array	Yes	Target URLs to extract from
prompt	string	Yes	Natural language description of the extraction task
schema	object	Yes	JSON Schema defining the output structure
agent	object	Yes	Agent configuration: {"model": "FIRE-1"}

Schema Definition

Define the structure of your expected output using JSON Schema:

{
  "type": "object",
  "properties": {
    "company_name": { "type": "string" },
    "pricing_plans": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "name": { "type": "string" },
          "price": { "type": "string" },
          "features": {
            "type": "array",
            "items": { "type": "string" }
          }
        },
        "required": ["name", "price"]
      }
    }
  },
  "required": ["company_name", "pricing_plans"]
}

Usage Examples

Python

from firecrawl import FirecrawlApp

app = FirecrawlApp(api_key="fc-YOUR-API-KEY")

result = app.extract(
    urls=["https://example.com/pricing"],
    prompt="Extract all pricing plans including plan name, monthly price, and features list",
    schema={
        "type": "object",
        "properties": {
            "plans": {
                "type": "array",
                "items": {
                    "type": "object",
                    "properties": {
                        "name": {"type": "string"},
                        "monthly_price": {"type": "string"},
                        "features": {
                            "type": "array",
                            "items": {"type": "string"}
                        }
                    },
                    "required": ["name", "monthly_price"]
                }
            }
        },
        "required": ["plans"]
    },
    agent={"model": "FIRE-1"}
)

for plan in result['data']['plans']:
    print(f"{plan['name']}: {plan['monthly_price']}")
    for feature in plan.get('features', []):
        print(f"  - {feature}")

Node.js

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

const app = new FirecrawlApp({ apiKey: "fc-YOUR-API-KEY" });

const result = await app.extract({
  urls: ["https://example.com/pricing"],
  prompt: "Extract all pricing plans including plan name, monthly price, and features list",
  schema: {
    type: "object",
    properties: {
      plans: {
        type: "array",
        items: {
          type: "object",
          properties: {
            name: { type: "string" },
            monthly_price: { type: "string" },
            features: { type: "array", items: { type: "string" } }
          },
          required: ["name", "monthly_price"]
        }
      }
    },
    required: ["plans"]
  },
  agent: { model: "FIRE-1" }
});

console.log(result.data);

cURL

curl -X POST https://api.firecrawl.dev/v1/extract \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer fc-YOUR-API-KEY' \
  -d '{
    "urls": ["https://example.com/pricing"],
    "prompt": "Extract all pricing plans including plan name, monthly price, and features list",
    "schema": {
      "type": "object",
      "properties": {
        "plans": {
          "type": "array",
          "items": {
            "type": "object",
            "properties": {
              "name": {"type": "string"},
              "monthly_price": {"type": "string"},
              "features": {"type": "array", "items": {"type": "string"}}
            },
            "required": ["name", "monthly_price"]
          }
        }
      },
      "required": ["plans"]
    },
    "agent": {"model": "FIRE-1"}
  }'

Use Cases

Forum Thread Extraction

Extract all comments from a paginated forum:

result = app.extract(
    urls=["https://forum.example.com/thread/12345"],
    prompt="Extract all user comments from this forum thread, including username, timestamp, and comment text. Follow pagination to get all pages.",
    schema={
        "type": "object",
        "properties": {
            "comments": {
                "type": "array",
                "items": {
                    "type": "object",
                    "properties": {
                        "username": {"type": "string"},
                        "timestamp": {"type": "string"},
                        "text": {"type": "string"}
                    }
                }
            }
        }
    },
    agent={"model": "FIRE-1"}
)

Filtered Product Listings

Extract products after applying filters:

result = app.extract(
    urls=["https://store.example.com/products"],
    prompt="Filter products by category 'Electronics' and price range '$50-$200', then extract all visible product names, prices, and ratings",
    schema={
        "type": "object",
        "properties": {
            "products": {
                "type": "array",
                "items": {
                    "type": "object",
                    "properties": {
                        "name": {"type": "string"},
                        "price": {"type": "string"},
                        "rating": {"type": "number"}
                    }
                }
            }
        }
    },
    agent={"model": "FIRE-1"}
)

Tabbed Content

Extract data from multiple tabs on a page:

result = app.extract(
    urls=["https://example.com/product/specs"],
    prompt="Click through all specification tabs (Overview, Technical, Dimensions) and extract the content from each tab",
    schema={
        "type": "object",
        "properties": {
            "overview": {"type": "string"},
            "technical_specs": {"type": "object"},
            "dimensions": {"type": "object"}
        }
    },
    agent={"model": "FIRE-1"}
)

Cost

FIRE-1 pricing is non-deterministic -- costs vary based on:

• Number of browser actions required
• Complexity of the page interaction
• Number of pages navigated
• Amount of data extracted

Use the credit calculator to estimate costs for your specific use case. FIRE-1 requests are generally more expensive than standard extraction due to the advanced browser automation and AI planning involved.

Limitations

• Cloud only -- not available in self-hosted deployments
• Rate limited to 10 requests per minute
• Variable cost -- complex interactions consume more credits
• URL required -- unlike Spark-1, you must provide target URLs
• No web search -- cannot discover URLs autonomously

Related Pages

• Agent API (Spark-1) -- Autonomous research without URLs
• Extract API -- Standard extraction without browser automation
• AI Models -- Spark-1 Mini vs Pro comparison
• LLM Extract (Legacy) -- Schema-based JSON extraction on scrape
• Browser Sandbox -- Manual browser control

---

Back to: Features Overview -->