TinyFish

TinyFish is a web infrastructure platform built specifically for AI agents, providing four APIs — Search, Fetch, Browser, and Agent — that let models and automated pipelines interact with the live web. The company is led by co-founder and CEO Sudheesh Nair and targets both individual developers and enterprise teams that need reliable, authenticated, and dynamic web access at scale.

What It Is

TinyFish sits between AI models and the live web, solving the problem that most valuable web data is locked behind logins, dynamic rendering, and anti-bot protections that standard LLM web search cannot reach. Rather than returning cached or indexed results, TinyFish executes real browser sessions, navigates authenticated systems, and returns structured outputs. The platform bundles four products under a single API key: a Search API for fresh structured results, a Fetch API that renders pages and returns clean markdown or JSON, a Browser API for stealth Chromium sessions with sub-250ms cold starts, and an Agent API for multi-step web automation.

How the Four APIs Work Together

Each API addresses a different layer of web access complexity:

• Search returns fresh, browser-rendered results from dynamic pages in structured JSON — not cached index data.
• Fetch renders any URL in a real browser and strips it to clean, token-efficient markdown, HTML, or JSON for downstream models and pipelines.
• Browser provides stealth Chromium sessions with engine-level anti-bot bypass, persistent state, and a 60-minute session cap. The homepage states a sub-250ms cold start and 85% anti-bot pass rate.
• Agent handles multi-step workflows: navigating pages, filling forms, authenticating into sites, and returning structured results. TinyFish claims 89.9% accuracy on the Mind2Web benchmark for production web tasks.

Search and Fetch use no credits; Agent and Browser consume credits per step or per four minutes of session time respectively.

Integration and Deployment Model

TinyFish is MCP-native, meaning a single JSON config block connects it to Claude, Cursor, and any MCP-compatible client. It also ships a Python SDK, CLI, npm package, and direct REST API for custom integrations. The integrations page lists connections to n8n, AG2, Vercel, Dify, and the MCP Registry. All four products share one API key and one credit pool, eliminating the need to manage separate auth configs or billing dashboards across multiple vendors.

Enterprise Architecture and Reliability Claims

The enterprise page states TinyFish has executed 40M+ operations with 95% task success and 99.99% uptime. Enterprise features include ISO 27001 certification, SOC 2 compliance, enterprise SSO, audit logs, VPC deployment, encrypted credential handling, and a dedicated support engineer. The platform supports parallel execution, persistent browser state for multi-step authenticated workflows, and structured outputs with validated schemas for downstream systems. The enterprise page describes use cases including vendor portal automation (one energy company reportedly automated 1,643 monthly logins), competitive price monitoring, insurance quoting workflows, compliance monitoring, and QA test automation (one test platform reportedly processes 20M steps monthly).

Developer Community and Accelerator

TinyFish runs an accelerator program backed by Mango Capital with a stated $2M seed pool, running 9 weeks virtually with rolling admissions. The program targets developers building web agents, with application via a working public project rather than a pitch deck. Community activity visible on the homepage includes posts from developers building integrations with Dify, AG2, and Replit, as well as hackathon participants using the platform. The company also operates a referral program and publishes an open-source cookbook on GitHub with example integrations.