TinyFish takes on the groundwork required for AI agents to browse, research, and operate the web. Four capabilities ─ search (Search API), content retrieval from URLs (Fetch API), autonomous web operations involving form input and login (Agent API), and a cloud browser that gets past bot protection (Browser API) ─ are all callable with the same API key. Because rendering, residential proxies, and bot-protection handling are absorbed at the infrastructure layer, you can pull structured data out of dynamically changing pages and sites that require a login. It is provided by Tiny Fish, Inc. in the United States and launched in August 2025.
Key Features
- Search API ─ live search results as structured JSON: Returns search results rendered in a real browser as ready-to-use JSON. A free tier of 30 requests/minute and 500 requests/hour lets you try it without paying
- Fetch API ─ turn any URL into clean text: Renders the URL you specify and returns the content as Markdown, JSON, or HTML. It also comes with a free tier of 150 URLs/minute and 1,000 URLs/day
- Agent API ─ autonomous multi-step web operations: Handles logins, form input, and procedures spanning multiple pages on its own, then extracts the results as structured data. The company highlights its accuracy on the Mind2Web benchmark
- Browser API ─ a cloud browser that gets past bot protection: Lets you rent browser sessions in the cloud equipped with residential proxies and bot-protection handling, so you do not have to run a fleet of headless browsers yourself
- Integration via MCP, CLI, and SDKs: An official MCP server and CLI are provided, so MCP-compatible clients such as Claude, Cursor, Codex, and Windsurf can call it directly
Pricing
| Plan | Billing | What you get |
|---|---|---|
| Search | Free | 30 requests/minute, 500 requests/hour |
| Fetch | Free | 150 URLs/minute, 1,000 URLs/day |
| Agent | $0.016/step (pay-as-you-go, wallet-based) | 2 concurrent runs |
| Browser | $0.002/minute (pay-as-you-go, wallet-based) | 5 concurrent sessions |
| Enterprise | Contact sales | ISO 27001, SSO, audit logs, VPC deployment, 99.99% uptime SLA |
Pricing is current as of September 2026. Check the official site for the latest rates.
Agent and Browser draw down from a wallet you fund in advance, charging only for what you use. Concurrency limits can be raised under an Enterprise contract.
Pros & Cons
✅ Pros
- You no longer have to juggle separate contracts for search, scraping, and browser automation ─ billing and key management come down to one
- The free tiers for Search and Fetch are enough for a realistic trial, so you can confirm whether the data you actually want comes back before putting a purchase through approval
- Running proxies and defeating bot protection is no longer your job, which hands the maintenance cost of chasing site-side changes to someone else
- Thanks to MCP support, you can also choose to drive the web straight from a Claude or Cursor conversation without writing code
⚠️ Cons
- Because Agent and Browser are pay-as-you-go, monthly costs are hard to forecast for workloads whose run counts are unpredictable
- Only Search and Fetch have free tiers, so funding the wallet is required from the moment you start testing automated web operations
- New accounts are capped at 2 concurrent Agent runs and 5 concurrent Browser sessions, which means heavy-parallel use cases involve talking to sales from the outset
Comparison with Similar Services
| Service | How TinyFish differs |
|---|---|
| Exa | Centered on a search-focused API. TinyFish bundles web operations and browser automation alongside search |
| Firecrawl | Focused on page extraction. TinyFish claims broader coverage plus agent-driven interaction capabilities |
| Parallel | Built around a web search API. TinyFish positions itself ahead on coverage and interaction capabilities |
| Tavily | Centered on search and extraction APIs. TinyFish also covers operational workflows involving logins and bot protection |
Who Is It For
- Businesses that want to patrol e-commerce sites regularly to monitor prices and stock
- Teams that need to automatically gather competing quotes from multiple portals in fields such as insurance and finance
- Engineers automating routine data extraction from internal or external systems that require a login
- Development teams building research or monitoring pipelines who would rather not source search, retrieval, and operations separately
- Small teams currently losing time to running their own headless browsers and proxies
Summary
Pulling data off the web tends to tangle together four distinct technical problems: search, retrieval, operation, and browser operations. TinyFish gathers all four behind a single API key and shoulders the work of rendering and site-side defenses. A sensible order is to start with the free tiers of Search and Fetch, confirm that your target sites really return the data in the shape you want, and move on to Agent and Browser once you need automated operations. If heavy parallelism is part of your plan, it is worth asking about concurrency limits early.