Product Introduction

1. FIRE-1 is an AI-powered agent integrated into Firecrawl’s web scraping infrastructure, designed to automate browser interactions and navigate complex website structures for advanced data extraction. It simulates human-like browsing behaviors, such as clicking buttons, filling forms, and handling dynamic content, to access data that traditional scraping methods cannot retrieve. The agent operates as an adaptive middleware layer, using machine learning to analyze page elements and execute precise navigation sequences.
2. The core value of FIRE-1 lies in its ability to transform static data collection pipelines into dynamic workflows capable of extracting multi-page, interaction-dependent datasets. By automating complex navigation patterns, it eliminates manual intervention for scenarios like pagination, infinite scroll, and JavaScript-rendered content. This enables users to gather comprehensive datasets from modern web applications that employ anti-scraping measures or require authenticated access.

Main Features

1. FIRE-1 automates browser-level interactions, including button clicks, form submissions, and dynamic element handling, to access data behind interactive interfaces. It executes predefined or AI-generated navigation sequences to simulate user workflows, such as logging into portals or traversing multi-step filters.
2. The agent supports multi-page data extraction by automatically managing pagination, infinite scroll, and session persistence across navigation steps. It intelligently detects and interacts with page elements like "Next Page" buttons while maintaining context between successive page loads.
3. Integration with Firecrawl’s /scrape and /extract API endpoints allows users to combine raw data collection with schema-based extraction, using natural language prompts to guide FIRE-1’s navigation logic. This enables both broad crawling and targeted data parsing in a single workflow.

Problems Solved

1. Traditional web scrapers fail to extract data from JavaScript-heavy websites requiring user interactions, such as e-commerce platforms with filters, forums with login walls, or dashboards with real-time updates. FIRE-1 bypasses these limitations by replicating human browsing actions programmatically.
2. The product targets developers, data engineers, and researchers who need structured datasets from modern web applications with complex authentication flows, anti-bot mechanisms, or dynamically generated content.
3. Typical use cases include scraping product catalogs with pagination, extracting comments from social media threads requiring infinite scroll, and collecting financial data from authenticated portals with multi-step navigation.

Unique Advantages

1. Unlike conventional scraping tools limited to static HTML parsing, FIRE-1 combines headless browser automation with AI-driven decision-making to handle websites designed to block automated access. It adapts to layout changes and recovers from missing elements through fallback strategies.
2. The agent’s prompt-based configuration allows users to define navigation logic in natural language, reducing the need for manual DOM analysis or XPath/CSS selector engineering. This enables rapid adaptation to new websites without code changes.
3. Firecrawl’s infrastructure provides built-in proxy rotation, CAPTCHA bypass, and automatic retries, allowing FIRE-1 to operate at scale without additional anti-blocking configurations. This integration ensures high success rates for enterprise-level scraping tasks.

Frequently Asked Questions (FAQ)

1. How do I activate FIRE-1 for a scraping job? Include an agent object with "model": "FIRE-1" and a navigation prompt in your API request to the /scrape or /extract endpoint. For /extract, the prompt parameter in the request body automatically guides the agent.
2. What is the cost difference between FIRE-1 and standard Firecrawl scraping? During the preview phase, /scrape requests using FIRE-1 consume 150 credits per call, while /extract costs approximately 8x more than non-agent extraction. Use Firecrawl’s token calculator for precise estimates based on task complexity.
3. Are there rate limits for FIRE-1-enabled requests? Yes, both /scrape and /extract endpoints with FIRE-1 are limited to 10 requests per minute. This ensures stable performance during the experimental phase and prevents resource overuse.
4. Can FIRE-1 handle websites with constantly changing UI elements? The agent uses real-time DOM analysis and adaptive retry logic to recover from missing elements or layout changes. Users can enhance reliability by providing fallback instructions in the navigation prompt.
5. Does FIRE-1 work with the /extract endpoint for schema-based data collection? Yes, when using /extract, include the agent parameter with "model": "FIRE-1", and the endpoint will automatically apply the provided schema after navigating through the required interactions.