> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fastapps.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Set up your server

> Build your MCP server and expose widgets as tools to ChatGPT.

## How does ChatGPT apps work?

Your MCP server is the foundation of every ChatGPT App. It exposes tools that the model can call, enforces authentication, and packages the structured data plus component that the ChatGPT client renders inline. This guide walks through the core building blocks with examples in FastApps.

## Tool Architecture

```mermaid theme={null}
graph TD
    A[ChatGPT User] -->|Natural language| B[ChatGPT Model]
    B -->|Tool call| E[FastApps Server]
    E -->|Structured data| F[Widget UI]
```

## BaseWidget Class

Every FastApps tool must inherit from `BaseWidget`:

```python theme={null}
class MyWidget(BaseWidget):
    # Required attributes
    identifier = "my_widget"           # Unique identifier
    title = "My Widget Title"          # Display name
    input_schema = MyInputModel        # Pydantic model
    invoking = "Setting up widget…"    # Progress message
    invoked = "Widget ready!"          # Completion message
    
    # Optional attributes
    description = "Widget description" # Help text
    widget_accessible = True           # Allow component calls
    
    def execute(self, inputs, ctx):
        # Your business logic here
        return {"data": "processed"}
```

## Next Steps

Ready to start building?

<CardGroup cols={3}>
  <Card title="Tool Basics" icon="book" href="/server/basics/index">
    Learn the fundamentals
  </Card>

  <Card title="API Integration" icon="plug" href="/server/integration/index">
    Connect to external services
  </Card>

  <Card title="Advanced Patterns" icon="wand-magic-sparkles" href="/server/advanced/index">
    Complex business logic
  </Card>
</CardGroup>

Or jump straight to the [**Quick Start Guide**](/quickstart/index) for a complete walkthrough.