v1.0.0 — Production Ready

AI that drives
your browser

Autonomous agent that navigates, clicks, types, and extracts data from any website. Monitor production. Automate workflows. Ship faster.

curl -fsSL https://nikeGunn.github.io/yeti-agent/install.sh | bash
pip install yeti-agent uv add yeti-agent uvx yeti-agent
yeti-agent
$ yeti-agent open https://your-production-site.com
🏔️ Browser launched — connected via CDP

$ python monitor.py
🤖 Agent: Navigating to checkout page...
🤖 Agent: Filling test order form...
🤖 Agent: Verifying payment flow... ✓ All checks passed
📊 Result: 3 pages tested, 0 errors, 2.4s avg response

Everything you need for
browser automation

From simple scraping to complex multi-step workflows. Yeti Agent handles it all with zero configuration.

🧠

LLM-Powered Decision Making

Supports OpenAI, Anthropic, Google Gemini, Groq, Ollama, and more. The agent sees the page and decides what to do next — like a human would.

🔍

Production Monitoring

Schedule autonomous agents to continuously test critical user flows — checkout, login, search, payments. Get alerted before your users notice.

🛡️

Enterprise Security

Domain allowlists, data masking, proxy support, headless mode, and security watchdogs. Built for production environments with strict compliance requirements.

📊

Structured Data Extraction

Extract data into Pydantic models with type validation. Define your schema, point the agent at a page, get clean structured JSON back.

🔌

Custom Actions & Tools

Extend the agent with your own Python functions. Register custom tools that the LLM can invoke during automation — API calls, database queries, anything.

☁️

Cloud & Self-Hosted

Run locally with your own Chrome, connect to remote CDP endpoints, or use cloud browser sessions. Deploy anywhere — Docker, Kubernetes, Lambda.

🔄

Multi-Tab Orchestration

Agent autonomously manages multiple tabs. Opens links in new tabs, switches context, aggregates data across pages — just like a power user.

📱

MCP Integration

Works as a Model Context Protocol server for Claude Desktop. Also connects to external MCP servers to extend agent capabilities.

Async & Blazing Fast

Built on async Python with event-driven architecture. Watchdog services handle downloads, popups, dialogs, and crashes in the background.

Three lines to production

The fastest path from idea to running agent. Works with your existing stack.

Quick Start
Monitoring
Data Extraction
Custom Tools
from browser_use import Agent, Browser
from browser_use import ChatOpenAI

agent = Agent(
    task="Go to HackerNews and find the top 3 AI stories",
    llm=ChatOpenAI(model="gpt-4o"),
)
result = agent.run_sync()
print(result.final_result())
from browser_use import Agent, Browser, BrowserProfile
from browser_use import ChatAnthropic

# Production-safe browser config
profile = BrowserProfile(
    headless=True,
    allowed_domains=["your-app.com", "*.your-app.com"],
)

agent = Agent(
    task="""
    1. Go to https://your-app.com/login
    2. Login with test credentials
    3. Navigate to checkout, add an item
    4. Verify payment form loads correctly
    5. Report any errors or broken elements
    """,
    llm=ChatAnthropic(model="claude-sonnet-4-20250514"),
    browser=Browser(browser_profile=profile),
)

# Run on a cron schedule for continuous monitoring
result = await agent.run(max_steps=50)
from pydantic import BaseModel
from browser_use import Agent, ChatOpenAI

class ProductData(BaseModel):
    name: str
    price: float
    rating: float
    in_stock: bool

agent = Agent(
    task="Extract all products from the first page",
    llm=ChatOpenAI(model="gpt-4o"),
    output_model_schema=ProductData,
)

# Returns typed, validated data — not messy HTML
products = await agent.run()
from browser_use import Agent, Tools, ChatGoogle

tools = Tools()

@tools.action(description="Send Slack alert when issue found")
def send_alert(message: str, severity: str) -> str:
    # Your alerting logic here
    slack.post(channel="#alerts", text=f"[{severity}] {message}")
    return "Alert sent"

agent = Agent(
    task="Monitor checkout flow and alert on any failures",
    llm=ChatGoogle(model="gemini-2.0-flash"),
    tools=tools,
)

How companies use Yeti Agent

From startups to enterprises, teams use Yeti Agent to automate what humans shouldn't spend time on.

Monitoring

Production Website Monitoring

Run autonomous agents 24/7 that test checkout flows, login pages, and critical user journeys. Get alerted before customers complain.

Testing

End-to-End QA Automation

Replace brittle Selenium scripts with AI agents that adapt to UI changes. Write tests in plain English — the agent figures out the selectors.

Extraction

Competitive Intelligence

Extract pricing, product data, and market intelligence from competitor websites. Output directly to Pydantic models for clean data pipelines.

Automation

Workflow Automation

Automate multi-step workflows across web apps — form filling, report generation, data entry. Works with any website, no API needed.

Testing

Regression Detection

Deploy agents post-release to verify nothing broke. Smarter than screenshot diffing — agents understand context and intent.

Monitoring

SLA Compliance Checking

Verify partner/vendor portals meet SLA requirements. Agents navigate, measure response times, check content, and generate compliance reports.

How it works

Event-driven architecture with watchdog services. Each component manages isolated state.

📋
Your Task
Plain English
🧠
LLM
GPT / Claude / Gemini
🤖
Agent
Plans & Executes
🌐
Browser
CDP Protocol
📊
Results
Typed Data

Event Bus

Decoupled services communicate through an event bus. Watchdogs for security, popups, downloads, DOM, and crashes all run independently.

CDP Direct Access

Uses cdp-use for typed Chrome DevTools Protocol access. Full control over browser state, network, DOM, and JavaScript execution.

Pydantic v2 Core

All data models use Pydantic v2 with strict validation. Type-safe from browser state through LLM output to your application code.

Production-grade from day one

Security, observability, and reliability features built for teams that can't afford downtime.

8+
LLM Providers Supported
20+
Built-in Browser Actions
3
Python Versions (3.11-3.13)
9
Watchdog Services
🔐

Security First

Domain allowlists/blocklists, sensitive data masking, proxy support, permission handling, and configurable security watchdogs.

📈

Observability

Structured logging, telemetry, GIF generation of runs, conversation export, and integration with Laminar for distributed tracing.

🐳

Deploy Anywhere

Docker-ready, Kubernetes-compatible, cloud browser support. Run headless in CI/CD, serverless functions, or dedicated infrastructure.

Start automating in
under 60 seconds

One command to install. Three lines to run.

curl -fsSL https://nikeGunn.github.io/yeti-agent/install.sh | bash