TinyFish Cookbook: Empowering AI Agents with Web Data

TinyFish offers a suite of four powerful endpoints, each tailored for specific web interaction needs of AI agents. The **Search** endpoint (api.search.tinyfish.ai) provides fast, structured web search results in JSON format, optimized for LLM consumption and designed for rank stability. It's an ideal drop-in retrieval layer for any agent. The **Fetch** endpoint (api.fetch.tinyfish.ai) excels at retrieving clean markdown, JSON, or HTML from any URL, offering real full-browser rendering. It's perfect for reading specific pages and feeding their content to LLMs, acting as a native replacement for tools like Firecrawl or hand-rolled Playwright solutions. For more complex tasks, the **Agent** endpoint allows users to provide a URL and a natural-language goal, with the agent navigating, acting, and returning clean JSON. This is best suited for multi-step flows and structured data extraction. Finally, the **Browser** endpoint provides access to a fully managed cloud browser, allowing users to connect their own Playwright or Selenium scripts for hardcore custom agent development and scripting. While Search and Fetch are free with generous rate limits, Agent and Browser are metered services.

TinyFish offers a versatile integration ecosystem, allowing users to interact with its services through multiple channels. The primary method is via its **REST API**, which can be accessed using simple cURL commands for Search, Fetch, and Agent functionalities. For users of AI assistants like Claude, Cursor, and ChatGPT desktop, TinyFish provides an **MCP Server** endpoint (mcp.tinyfish.ai) that can be dropped into their client configurations for seamless integration. Developers can also utilize the **Command Line Interface (CLI)**, installed via `npm install -g @tiny-fish/cli`, for direct interaction with search and fetch commands, writing results to the filesystem to conserve token usage. For programmatic integration within applications, TinyFish offers robust **SDKs** for both Python (`pip install tinyfish`) and TypeScript (`npm install @tiny-fish/sdk`). These SDKs provide full parity with all TinyFish endpoints, including Search, Fetch, Browser, Agent, and Vault, making it easy to incorporate TinyFish capabilities into existing codebases.

Beyond the featured live demos, the TinyFish Cookbook organizes a wide array of recipes into specialized categories, catering to diverse use cases: **AABW Vietnam Hackathon Samples**: Projects developed during the Agentic AI Build Week hackathon, focusing on tracks like Mobility, Aviation, Financial Services, Retail, and F&B. **Shopping & Deals**: Recipes like `bestbet` for sports betting odds comparison, `game-buying-guide` for video game pricing, `lego-hunter` for rare Lego inventory, `openbox-deals`, `waifu-deal-sniper` for anime figure collectors, and `wing-command` for tracking chicken wing deals. **Travel, Stays & Local**: Includes `stay-scout-hub` for cross-site lodging search, `viet-bike-scout`, `district-rent-shark` for Vietnamese rental market intelligence, `restaurant-comparison-tool` for safety and allergen information, and `saigon-happy-hour-sniper`. **Research & Market Intelligence**: Features `research-sentry` for academic research, `silicon-signal`, `competitor-analysis` for pricing intelligence, `competitor-scout-cli` for competitor feature research, `logistics-sentry` for port congestion tracking, and `tenders-finder` for Singapore government tenders. **Education & Discovery**: Offers `tutor-finder`, `summer-school-finder`, and `scholarship-finder` for educational resource discovery. **Developer Tools**: Includes `code-reference-finder` for code snippet examples, `fast-qa` for no-code QA testing, and `tinyskills` for generating developer guides. **Finance & Decisioning**: Provides `loan-decision-copilot` for loan comparison. **Lifestyle & Health**: Features `anime-watch-hub` for media discovery and `pharmacy-panic` for medicine price comparison. New recipes are added weekly, encouraging continuous community engagement.

Embarking on your journey with TinyFish is straightforward. The initial step involves obtaining an API key by signing up at agent.tinyfish.ai; no credit card is required, and the Search and Fetch endpoints are immediately available and free. Once you have your API key, you can begin running applications. The cookbook provides practical examples for using the **cURL** command-line tool, demonstrating how to send requests to the Agent endpoint for tasks like extracting subscription plans and prices. For developers, **Python** and **TypeScript** code snippets are also provided, showcasing how to make POST requests to the Agent endpoint, handle streaming responses, and integrate TinyFish into their projects. For quick local sharing of demos, TinyFish offers a free and simple tool called `tinyfi.sh`.