TinyFish has expanded its web infrastructure platform from a single web agent to a four-product API suite, now available for developers building production workflows with AI coding agents. The platform offers four endpoints under one API key:

Web Agent: Enables autonomous multi-step workflow execution on live websites.
Web Search: Returns structured JSON results via a custom Chromium engine with approximately 488ms P50 latency.
Web Browser: Provides managed stealth Chrome sessions via CDP with sub-250ms cold start and 28 anti-bot mechanisms built at the C++ level.
Web Fetch: Converts any URL to clean Markdown, HTML, or JSON with full browser rendering.

The platform is publicly available with 500 free steps and no credit card required on signup.

The most developer-facing aspect of this release is the TinyFish CLI and Agent Skill. The CLI is installable via npm install -g @tiny-fish/cli allows any terminal or AI coding agent direct access to all four endpoints without SDK wiring or manual configuration. The Skill is a markdown instruction file that educates coding agents such as Claude Code, OpenClaw, Codex, Cursor, and OpenCode on when and how to call each endpoint. Once installed, the agent gains the entire TinyFish capability set without needing a single line of integration code. According to TinyFish benchmarks, the CLI-plus-Skill approach reduces tokens per operation by 87 percent compared to native MCP flows and achieves twice the task completion rate on complex multi-step tasks, primarily because output writes to the filesystem rather than flooding the agent context window.

0:00

/0:49

TinyFish

The rationale behind this architecture addresses a recurring failure mode in agent-based web workflows. Most native web fetch tools in coding agents retrieve the entire page, including CSS, scripts, navigation bars, footers, and ads, before reaching the needed content. TinyFish Fetch renders the page in a full browser, removes irrelevant elements, and returns clean content as Markdown or JSON. A related issue arises when teams use separate providers for search, fetch, and browser automation: session fingerprints diverge, context breaks across steps, and integration work increases with each additional tool. TinyFish built all four layers in-house and offers them as a unified platform, allowing the system to trace a task from search through fetch to browser execution and apply that signal back across every layer as a compounding data loop. The company claims 89.9 percent web agent accuracy across 300 tasks on the Mind2Web benchmark, ranking first in that evaluation.

Test TinyFish yourself!

Take me there!

TinyFish is backed by $47 million in Series A funding led by ICONIQ and counts Google, DoorDash, Cigna, Grubhub, Volkswagen, and NEC among its enterprise clients. The company has processed over 40 million agent operations and reports 99.99 percent platform uptime. The primary audience for this launch is developers building B2B tooling, internal copilots, data extraction pipelines, and any production AI workflow that requires reliable, live web access at scale.

TestingCatalog

Message History

Delivery option(s)