Works toward the outcome
Give TinyAgent the result you need. It moves through the website, observes what changed, adjusts the next step, and continues until the goal is complete.
/agent
Give TinyAgent the goal.
It finishes the work.
Navigates live websites, handles logged-in sessions, adapts when the path changes. Returns structured results your agent can actually use.
Get a liability quote from kestrel.com/quote. Consulting firm, five employees, California.
Your agent
TinyAgent · kestrel.com/quote
Opened quote formkestrel.com
Filled business details5 employees · CA
Selected general liability$1M limit
Submitted for underwritingwaited 8.2s
Extracted quote$1,284 / yr
$1,284 a year with Kestrel. $1M limit, $500 deductible. Nobody wrote a script.
One goal in. Browser work happens outside your agent’s context. Structured results come back.
Trusted by
Use cases
What TinyAgent makes possible.
Use TinyAgent when the workflow cannot be solved by reading one page. It is for web work where the path is dynamic, authenticated, multi-step, or unknown in advance.
01
Internal automations
Automate the web tasks your team still does by hand: checking statuses, updating forms, pulling records, reconciling information, and moving work across tools that do not expose clean APIs.
02
Authenticated data workflows
Work inside dashboards, account pages, vendor portals, customer systems, marketplaces, and private tools using persistent browser profiles.
03
Exploration and research
Send agents into websites where the path is not known in advance. TinyAgent can search within a site, compare options, open relevant paths, reject dead ends, and keep going until it finds the answer.
04
Agentic automation at scale
Run browser-based workflows in parallel across accounts, sites, records, filings, products, or locations. Each run works toward its own goal.
05
Human-judgment workflows
Use TinyAgent when the task requires reading, choosing, filtering, comparing, and deciding what to do next, not just extracting fields from a known URL.
Why TinyAgent
Built for goal completion on the live web.
Has its own profiles
Persistent browser profiles keep sessions and cookies available across runs, so TinyAgent can return to portals and dashboards like a known user instead of a blank browser.
Runs in parallel
Run TinyAgent across many accounts, sites, records, filings, products, or locations at once. Each browser run works independently toward its own goal and returns structured output.
Inspect every run
Every run gives your team the result plus the trail behind it: structured output, logs, screenshots, and source URLs. Your agent can use the structured result directly, while your team can inspect the full run when needed.
Use it anywhere
Use TinyAgent wherever your agents work.
Start from a goal in the CLI, MCP, SDK, API, or your own automation stack.
TypeScript SDK
Skills
LangChain
REST API
Codex
Hermes Agent
n8n
MCP
Cursor
OpenClaw
Dify
Python SDK
Claude Code
CLITypeScript SDKSkillsLangChainREST APICodexHermes Agentn8nMCPCursorOpenClawDifyPython SDKClaude CodeCLIREST APIPython SDKOpenClawCodexClaude CodeSkillsCursorTypeScript SDKHermes AgentCLIn8nLangChainMCPDifyREST APIPython SDKOpenClawCodexClaude CodeSkillsCursorTypeScript SDKHermes AgentCLIn8nLangChainMCPDifyHermes AgentMCPTypeScript SDKSkillsREST APICodexLangChainCLIn8nClaude CodePython SDKCursorDifyOpenClawHermes AgentMCPTypeScript SDKSkillsREST APICodexLangChainCLIn8nClaude CodePython SDKCursorDifyOpenClawProof
Accuracy you can build production workflows on.
When a workflow has many steps, every decision matters. TinyAgent is designed to complete the goal reliably across dynamic, authenticated, and multi-step websites.
1TinyFish91.1%
2BrowserUse88.3%
3Smooth86.6%
4Notte84.2%
WebVoyager · 641 tasks, 15 live websites · independent evaluation
1TinyFish89.9%
2Gemini 2.5 Computer Use69.0%
3OpenAI Operator61.3%
4Claude Computer Use56.3%
Online-Mind2Web · live task benchmark
PricingSimple pricing
Simple pricing
for agent work.
$0.016
per agent step
You pay for the agent work performed. TinyFish handles token costs, browsers, storage, retries, proxies, and run infrastructure.
FAQ
Questions, answered.
What is TinyAgent, and is it a web agent API?
TinyAgent is a web agent API for programmatic workflows that need to complete goals on live websites. It moves through pages, handles multi-step tasks, works with authenticated sessions, and returns structured results, logs, screenshots, and source URLs.
When should I use TinyAgent instead of TinyFetch?
Use TinyFetch when you have a URL and need clean page content. Use TinyAgent when the task requires clicking, searching within a site, logging in, filling forms, comparing options, or working through a sequence of pages.
Can TinyAgent work behind login?
Yes, in two ways. Persistent browser profiles keep saved sessions and cookies available across runs, so TinyAgent can return to a portal or dashboard already signed in. Where a fresh sign-in is needed, TinyVault integration lets the agent log in without exposing credentials to the LLM.
What are profiles?
Profiles work like a Chrome profile for your agent. They preserve logged-in context across runs, so the agent can return to the same portal, dashboard, marketplace, or account page without starting from a blank browser every time.
Can I run many TinyAgent tasks at once?
Yes. TinyAgent can run browser-based workflows in parallel across accounts, sites, records, filings, products, or locations.
How is TinyAgent priced?
TinyAgent is $0.016 per step, drawn from your Wallet funds. There is no subscription and no monthly minimum.
What does TinyFish handle behind the scenes?
TinyFish handles LLM inference, browser infrastructure, retries, proxies, storage, and run infrastructure. Your agent receives structured output for downstream reasoning, while logs, screenshots, and source URLs stay available for observability.
Turn web goals
into finished work.
Use TinyAgent for workflows that need judgment, persistence, authenticated sessions, and action across live websites.
No credit card required