# TinyFish — Web APIs and browser agents for search, fetch, and multi-step automation, priced pay-as-you-go through one shared API key.

> Source: The Agents Index — https://theagentsindex.com/tinyfish (structured, researched, re-verified)
> Facts last verified: 2026-08-17

TinyFish is a web automation company offering five products under a single API key: Mako, a web-native model for live workflows, plus Search, Fetch, Agent, and Browser. It positions itself as infrastructure for pulling structured data from live web pages and running multi-step browser tasks, integrating natively with MCP clients like Claude and Cursor and with automation platforms such as n8n and Dify. It serves developers, competitive-intelligence teams, and enterprises that need programmatic access to sites that resist scraping.

| Fact | Value |
| --- | --- |
| Website | https://tinyfish.ai |
| Pricing | Search and Fetch are free; Agent runs $0.016 per step and Browser $0.002 per minute through a prepaid Wallet ($10 minimum deposit) |
| API | Yes |
| Best for | Developers who need free structured web data plus paid browser automation on one key |
| Not for | Teams that need airtight CAPTCHA bypass out of the box: TinyFish still routes the toughest CAPTCHAs to third-party solvers like CapSolver, adding extra tooling and cost for sites with aggressive bot defenses. |

## Pricing

| Tier | Price |
| --- | --- |
| Free (Search & Fetch) | $0 / per request |
| Wallet (Agent & Browser, pay-as-you-go) | $0.016/step (Agent), $0.002/min (Browser) / usage-based |
| Enterprise | Custom (contact sales) |

## Verdict

TinyFish bundles five web-data products, Mako, Search, Fetch, Agent, and Browser, under a single API key and a prepaid Wallet instead of a subscription. The free Search and Fetch tier is a real on-ramp: no card, 150 URLs and 30 searches per minute. Where it gets interesting is Browser and Agent, priced per minute and per step, with anti-bot numbers (85% pass rate, 99.3% detection coverage) aimed squarely at sites that block scrapers. The catch is CAPTCHA handling still leans on outside tools like CapSolver, and pricing reporting is inconsistent across sources, so teams should verify current rates directly with TinyFish before committing budget.

## How it works

1. Get one API key that covers Mako, Search, Fetch, Agent, and Browser.
2. Call Search or Fetch for free to pull structured JSON or clean markdown/HTML from a URL or search query.
3. Fund a prepaid Wallet ($10 minimum) to run Agent for multi-step automation ($0.016/step) or Browser for cloud sessions ($0.002/min).
4. Watch runs live via Live Execution Streaming, or trigger them through an MCP client like Claude/Cursor or a no-code node in n8n/Dify.

## Who it's for

- Developers who need clean, structured content pulled from arbitrary URLs for a data pipeline
- Competitive intelligence and pricing teams monitoring competitor sites at scale
- Engineering teams automating multi-step web tasks (portal checks, form fills) on sites with anti-bot defenses
- No-code builders using n8n or Dify who want a browser agent node without writing API calls

## Strengths and weaknesses

- ✓ Search and Fetch are genuinely free with no card required, so teams can prototype data pulls before paying for Agent or Browser usage.
- ✓ Usage-based Wallet pricing ($0.016/step Agent, $0.002/min Browser) avoids locking small teams into a fixed monthly plan.
- ✓ Reports strong anti-bot numbers (85% pass rate, 99.3% detection coverage) and sub-250ms Browser cold starts for scraping-resistant sites.
- ✓ MCP-native from the start, plugging directly into Claude and Cursor plus community nodes for n8n and Dify.
- ✗ CAPTCHA handling on the most aggressive sites still needs a third-party solver like CapSolver, so the anti-bot story has a real gap.
- ✗ Independent pricing coverage conflicts with TinyFish's own pricing page (one review cites $15 and $150 monthly plans, while the vendor site describes pure pay-as-you-go Wallet pricing), making budgeting hard to pin down before signup.
- ✗ Default concurrency is capped (2 concurrent Agent runs, 5 Browser sessions), so higher throughput requires contacting sales.
- ✗ Headline performance numbers (98.7% accuracy, 89.9% Mind2Web score) come from the vendor or a single review site, not an independently reproduced benchmark.
- ⚠ Binary files like images and video aren't supported by Fetch; only text, PDF, and JSON content extract cleanly.
- ⚠ Independent coverage is thin, one third-party review (aivorapulse.com) alongside the vendor's own site and posts, so claims are not yet broadly corroborated.

## Key features

- **Fetch API** — Converts any URL into clean markdown, HTML, or JSON, batching up to 10 URLs per request with CSS selector scoping, conditional requests, and cache TTL control.
- **Search API** — Returns fresh, browser-rendered structured JSON results from the live web instead of a stale index, free at 30 requests per minute.
- **Agent** — Runs multi-step web automation (navigating, dismissing pop-ups, extracting data) billed per step, scaling from 2 concurrent runs by default to 1,000 for enterprise.
- **Browser** — Cloud browser sessions with sub-250ms cold starts, an 85% anti-bot pass rate, and 99.3% detection coverage, priced per minute.
- **Live Execution Streaming** — Lets a user watch an agent move through a website step by step in real time instead of waiting on a final result.
- **MCP-native integration** — Works directly with Claude, Cursor, and any MCP-compatible client, plus community nodes for n8n and Dify and a listing on the MCP registry.

## Use cases

- **Competitor price monitoring** — An open-source TinyFish agent generates or accepts competitor pricing URLs, then extracts tiers, overage costs, and feature comparisons into structured JSON in under two minutes per competitor.
- **Portal and status checks at scale** — A team runs TinyFish Agent to complete 50 portal checks in under 2.5 minutes as part of a production monitoring workflow.
- **Clean content for LLM pipelines** — A developer calls Fetch on a batch of URLs to get markdown or JSON text back, scoped to specific page sections, for feeding into a downstream model.
- **No-code web automation** — A non-developer builds a workflow in n8n or Dify using TinyFish's community node to trigger a browser agent without writing API calls.

## Integrations

Claude (MCP) · Cursor (MCP) · n8n · Dify · MCP registry

## Sources

- https://tinyfish.ai
- https://docs.tinyfish.ai/fetch-api
- https://tinyfish.ai/integrations
- https://tinyfish.ai/pricing
- https://www.youtube.com/watch?v=YDWNz2XzLTU
- https://aivorapulse.com/tinyfish-ai-review-2026-guide/
- https://www.linkedin.com/posts/tinyfish-ai_your-competitor-dropped-prices-40-last-week-activity-7423172258397085696-EVuy