Skip to content

Built-in Agents

The toolkit includes three reference agents demonstrating different patterns and capabilities.

Echo Agent

A deterministic agent that doesn't use AI models. Perfect for testing webhook delivery and SSE streaming.

Usage

uv run layercode-create-app run --agent echo

Behavior

  • Sends a welcome message when the call starts
  • Echoes back whatever the user says
  • Doesn't require API keys for AI providers

Use Cases

  • Testing webhook configuration
  • Verifying tunnel connectivity
  • Understanding the basic agent flow
  • Development without AI costs

Implementation

Location: src/layercode_create_app/agents/echo.py

@agent("echo")
class EchoAgent(BaseLayercodeAgent):
    async def process(self, event):
        if event.type == "call_started":
            return "Welcome! I'll echo everything you say."
        elif event.type == "transcript":
            return f"You said: {event.transcript}"

Starter Agent

A general-purpose conversational assistant with progressive disclosure and automatic transcription cleanup.

Usage

uv run layercode-create-app run --agent starter

Features

  • Clean, natural responses
  • Context-aware conversations
  • Automatic transcription cleanup
  • Handles interruptions gracefully

Configuration

Requires an AI provider API key:

OPENAI_API_KEY=sk-...
# OR
GOOGLE_GENERATIVEAI_API_KEY=...

Model Selection

# Use default model from .env
uv run layercode-create-app run --agent starter

# Override with specific model
uv run layercode-create-app run --agent starter --model openai:gpt-4

Use Cases

  • Customer support
  • General Q&A
  • Information lookup
  • Friendly conversation

Prompt

Location: src/src/layercode_create_app/agents/prompts/starter.txt

The prompt emphasizes: - Concise responses - Progressive disclosure - Natural conversation flow - Handling of speech artifacts

Bakery Agent

A persona-driven agent demonstrating custom tools for business operations.

Usage

uv run layercode-create-app run --agent bakery --tunnel

Features

  • Bakery persona with character
  • Custom tools for business logic:
  • Menu lookup
  • Order placement
  • Reservation booking
  • Demonstrates PydanticAI tool calling

Tools

The bakery agent has three tools:

1. Get Menu 

@tool
async def get_menu(ctx: RunContext) -> str:
    """Retrieve the bakery menu"""
    return menu_items

2. Place Order 

@tool
async def place_order(
    ctx: RunContext,
    items: list[str],
    quantity: list[int]
) -> str:
    """Place an order for bakery items"""
    # Order processing logic

3. Make Reservation 

@tool
async def make_reservation(
    ctx: RunContext,
    date: str,
    time: str,
    party_size: int
) -> str:
    """Book a table reservation"""
    # Reservation logic

Use Cases

  • Restaurant ordering
  • Business automation examples
  • Learning tool implementation
  • Custom persona development

Prompt

Location: src/src/layercode_create_app/agents/prompts/bakery.txt

Features a friendly bakery owner persona with knowledge of the menu and operations.

Comparing Agents

Feature Echo Starter Bakery
AI Required No Yes Yes
Custom Tools No No Yes
Persona None Generic Bakery Owner
Use Case Testing General Business Logic
Complexity Simple Medium Advanced

Running Multiple Agents

You can run multiple agents simultaneously on different ports:

# Terminal 1
uv run layercode-create-app run --agent echo --port 8000

# Terminal 2
uv run layercode-create-app run --agent starter --port 8001

# Terminal 3
uv run layercode-create-app run --agent bakery --port 8002

Next Steps