TinyFish

Product Introduction

1. Definition: TinyFish is a unified, serverless API platform for AI web agents and applications. It provides enterprise-grade infrastructure for web search, content extraction, browser automation, and multi-step web interaction.
2. Core Value Proposition: TinyFish exists to solve the operational complexity and unreliability of accessing live, dynamic, and authenticated web data for AI systems. It enables developers to build AI agents that can deliver accurate, current information and automate complex web workflows at scale through a single, reliable API.

Main Features

1. Search API: A real-time web search engine that returns structured JSON results from dynamically rendered pages. Unlike traditional cached search APIs, it uses real browser instances to execute searches, ensuring access to the most current information from sources like news sites, financial data, and rapidly updating platforms. It is designed for use cases like pricing intelligence, earnings monitoring, and breaking news aggregation.
2. Fetch API: A content extraction tool that renders any URL in a headless browser and returns clean, token-efficient data in Markdown, JSON, or HTML formats. This process strips away navigation, ads, and other boilerplate, delivering only the core content as it appears to a human user. It is essential for research pipelines, content monitoring, and feeding clean data into large language models (LLMs).
3. Web Agent API: A multi-step web automation engine designed to navigate pages, interact with forms, authenticate into websites, and extract structured data. Benchmarked at 89.9% accuracy on the Mind2Web evaluation suite, it uses semantic understanding of page elements instead of brittle CSS selectors, making it robust against website changes. It is built for production tasks like insurance quoting, travel booking, and data entry automation.
4. Browser API: Provides persistent, stealth cloud browser sessions capable of bypassing anti-bot protections like Cloudflare and Akamai. It maintains authenticated state, handles dynamic JavaScript rendering, and offers sub-250ms cold starts. This is critical for accessing secure vendor portals, internal dashboards, and other gated web applications where traditional scraping fails.

Problems Solved

1. Pain Point: AI agents and applications struggle with accessing reliable, real-time data from the modern web, which is dynamic, JavaScript-heavy, and often protected by anti-bot measures. Traditional web scraping is brittle, slow, and cannot handle authenticated sessions or complex interactions.
2. Target Audience: AI/ML Engineers building data pipelines, Product Managers for competitive intelligence tools, Developers of AI agent startups, Data Scientists requiring live web data for models, and Automation Engineers tasked with web workflow automation.
3. Use Cases:
   • Competitive Price Monitoring: Automatically tracking competitor pricing on e-commerce sites.
   • Insurance Quoting Automation: Logging into multiple carrier portals, filling forms, and extracting structured quotes.
   • Real-time Market Intelligence: Gathering fresh financial data, news, and social sentiment.
   • Social Listening: Extracting clean content from forums, review sites, and social media for analysis.
   • Autonomous QA Testing: Using AI agents to perform automated user journey testing on web applications.

Unique Advantages

1. Differentiation: Unlike stitching together separate vendors for search, scraping, and browser automation, TinyFish offers a unified platform. This means one API key, one credit pool, one SDK, and intelligent routing that automatically selects the right tool (Search, Fetch, Browser, Agent) for each step of a workflow without developer intervention.
2. Key Innovation: The platform's "semantic element finding" for its Web Agent, which allows it to understand page purpose and interact with elements based on their function rather than fragile locators. Combined with its serverless, massively parallel architecture for running hundreds of browser sessions simultaneously, this enables reliable automation at a scale previously difficult to achieve.

Frequently Asked Questions (FAQ)

1. How does TinyFish handle websites with anti-bot protection like Cloudflare? TinyFish's Browser API uses advanced fingerprinting techniques and engine-level stealth configurations to mimic human browser behavior, achieving a high bypass rate for common anti-bot systems, allowing access to protected sites.
2. What is the difference between TinyFish Search and Google Search API? TinyFish Search renders pages in a real browser, returning live data from dynamic sites (e.g., live pricing, recent tweets). Google's API primarily returns cached, static results from its index, which can be outdated for fast-changing information.
3. Can I use TinyFish for large-scale web scraping projects? Yes, TinyFish is built for production-scale workloads. Its serverless architecture can run hundreds of concurrent sessions, and its unified credit system simplifies billing for mixed workloads involving search, fetching, and browser automation.
4. How does the TinyFish Web Agent achieve high accuracy on complex tasks? It combines computer vision and LLM-based semantic understanding to interpret web pages contextually, deciding on actions like "click the login button" or "fill the address field" rather than relying on code selectors that break when a website's design changes.
5. Is there a free tier for the TinyFish API? Yes, TinyFish offers a free tier that includes access to its Search and Fetch APIs without consuming credits, allowing developers to test real-time search and content extraction capabilities.