TinyFish

import { tinyfish } from "scrape-sdk/tinyfish";

const scraper = tinyfish({
  apiKey: process.env.TINYFISH_API_KEY!,
});

TinyFish uses separate REST surfaces for Fetch and Search. Fetch renders a URL and returns clean Markdown or HTML; Search returns structured ranked results. Both are free within the account limits (currently 150 URLs/minute for Fetch and 30 requests/minute for Search).

const page = await scraper.scrape("https://example.com");
const hits = await scraper.search("web automation tools", {
  location: "US",
  language: "en",
  purpose: "Find primary sources for an agent answer",
});

The goal-based Agent surface is metered, so it is disabled by default. Enable it explicitly when the agent needs interaction or structured extraction from a multi-step flow:

const scraper = tinyfish({
  apiKey: process.env.TINYFISH_API_KEY!,
  enableAgent: true,
});

const result = await scraper.agent("https://example.com", {
  goal: "Find the pricing page and return the first three plan names and prices.",
  schema: {
    type: "object",
    properties: { plans: { type: "array" } },
    required: ["plans"],
  },
});

fromEnv({ tinyfishAgent: true }) enables the same paid surface. TinyFish does not provide native map() or crawl() in this adapter, and target-page headers, custom wait times, and onlyMainContent: false are rejected rather than silently ignored.