Skip to main content
When a user calls one of your purchased phone numbers, the call is received by Daily’s infrastructure and placed on hold. This triggers a webhook that initializes the Pipecat Bot. There are two ways to handle this initialization:
  1. Host the webhook yourself in your application infrastructure.
  2. Let Daily handle the webhook (coming soon).
For most use cases, letting Daily manage the webhook should be sufficient. However, if you need to customize the bot’s behavior—for example, based on the caller’s number or the number dialed—you’ll need to host the webhook yourself to apply that logic. We provide example servers built with Next.js and FastAPI to help you get started with your own implementation.

FastAPI Webhook Server

Python-based webhook server implementation using FastAPI for handling dial-in calls

Next.js Webhook Server

JavaScript-based webhook server implementation using Next.js for handling dial-in calls

​
Handling Dial-In Webhook (room_creation_api)

To support inbound calls, you currently need to host a server that handles incoming call webhooks. In the near future, Daily will support managing incoming calls directly and provide an endpoint—similar to {service}/start—to handle this automatically. When someone calls your purchased number, Daily sends a webhook request containing information about the call. Your server can use this data to take the call off hold and connect it to a Pipecat Bot. Here’s a sample of the webhook payload: 
{
  "To": "+1CALLER",
  "From": "+1PURCHASED",
  "callId": "string-contains-uuid",
  "callDomain": "string-contains-uuid"
}
Your webhook handler should then forward this data to the /start endpoint to initialize the session. Here are the required fields to set:
  • Set createDailyRoom to true
  • Configure dailyRoomProperties with the following sip settings:
    • sip_mode: “dial-in”
    • num_endpoints: 1 (set to 2 for call transfers)
    • display_name: typically set to the From number
  • In the body, map the webhook payload to dialin_settings and convert camelCase to snake_case (e.g., callId → call_id).
curl --request POST \
--url https://api.pipecat.daily.co/v1/public/{service}/start \
--header 'Authorization: Bearer $API_KEY' \
--header 'Content-Type: application/json' \
--data-raw '{
    'createDailyRoom': true,
    'dailyRoomProperties': {
        'sip': {'display_name': '+1CALLER', 'sip_mode': 'dial-in', 'num_endpoints': 1},
        'exp': 1742353314
    },
    'body': {
        'dialin_settings': {
            'from': '+1CALLER',
            'to': '+1PURCHASED',
            'call_id': 'callid-uuid',
            'call_domain': 'domain-uui'
        }
    }
}

​
Receiving a Call on the Bot

The {service}/start endpoint creates a Pipecat bot instance and forwards the incoming request to it. Within the bot, pass the dialin_settings from the request body to the DailyTransport, see example below. The incoming PSTN or SIP call will then be automatically routed to the bot. 
transport = DailyTransport(
    room_url,
    token,
    "Voice AI Bot",
    DailyParams(
        api_key=os.getenv("DAILY_API_KEY"),  # Required for dial-in
        audio_in_enabled=True,
        audio_out_enabled=True,
        dialin_settings=dialin_settings,     # Received from the {service}/start body
        vad_analyzer=SileroVADAnalyzer(),
    ),
)
Since this is an inbound call, the bot should begin speaking as soon as the remote user joins the call. 
 @transport.event_handler("on_first_participant_joined")
    async def on_first_participant_joined(transport, participant):
        if dialin_settings:
            # the bot will start to speak as soon as the call connects
            await task.queue_frames([LLMRunFrame()])

​
Next Steps

After setting up your webhook server, you can implement a Pipecat bot that handles the call interactions:

Dial-in/Dial-out Bot Example

Complete example of a Pipecat bot implementation that handles both incoming (dial-in) and outgoing (dial-out) calls
 This starter template provides all the necessary components to build a production-ready telephony bot.