Test Presentation Page

Welcome to Vanna 2.0

A production-grade framework for building conversational AI agents

Use arrow keys or swipe to navigate →

👤 User-Aware from Day One

• Every request has an authenticated user
• Permissions enforced automatically
• Group-based access control
• Multi-tenancy built-in

Quick Start Code

from vanna import Agent
from vanna.tools import QueryTool

# Create your agent
agent = Agent(
    tools=[QueryTool()],
    llm="claude-sonnet-4"
)

# Run a query
response = agent.run("Show me sales by region")

Key Features

• User-aware execution

Key Features

• User-aware execution
• Permission-based tools

Key Features

• User-aware execution
• Permission-based tools
• Rich UI components

Key Features

• User-aware execution
• Permission-based tools
• Rich UI components
• Composable architecture

Key Features

• User-aware execution
• Permission-based tools
• Rich UI components
• Composable architecture
• Production-ready

Architecture Diagram

🔐 Permission-Based Tools

class SensitiveTool(Tool):
    @property
    def required_permissions(self):
        return ["admin", "write_data"]

    async def execute(self, context, args):
        # Only executes if user has permissions
        return ToolResult(success=True)

🎨 Rich UI Components

• Stream interactive components, not just text
• DataFrames, charts, status cards
• Progress bars and real-time updates
• Dashboard-quality experiences

Composable Architecture

Composable Architecture

Ready to Get Started?

Build your first agent in 5 minutes

View Quickstart →