Phase 6: Real-Time Transport & Telephony · 50 min · Python · Telnyx SDK · Vonage SDK

SIP, PSTN, and Enterprise Phone System Integration

SIP is the protocol that runs the world's phone systems. If you deploy enterprise voice agents, you need it.

Hiring signal: SIP and enterprise telephony knowledge distinguishes voice AI consultants and FDEs from generalists.

What you will learn

Understand SIP signaling, SDP, and RTP for enterprise telephony
Implement SIP trunking for enterprise PBX integration
Compare Telnyx, Vonage, and AWS Connect for enterprise deployments
Evaluate on-premise SIP options: Asterisk, FreeSWITCH

The Problem

Large organizations have existing phone infrastructure — Cisco PBXs, Avaya systems, SIP trunks. They won't switch to Twilio. Your voice agent needs to integrate with their SIP infrastructure, receive calls from their existing numbers, and route audio through their systems.

The Concept

What is SIP?

SIP (Session Initiation Protocol) is the signaling protocol for VoIP calls — it sets up, manages, and tears down voice sessions. It's like HTTP but for phone calls.

SIP Component	Role	Analogy
SIP Registrar	Registers user locations	DNS for phones
SIP Proxy	Routes calls	Router for calls
SIP PBX	Manages extensions, routing	Switchboard
SIP Trunk	Connects PBX to PSTN	Phone line
SDP (in SIP)	Negotiates audio codecs	Menu of options

SIP Call Flow

SIP vs Twilio

Factor	SIP (Direct)	Twilio
Infrastructure	Need SIP server/PBX	Twilio handles it
Cost per minute	$0.001-0.005	$0.005-0.015
Audio quality	Configurable (G.711, G.722, Opus)	μ-law 8kHz default
Setup complexity	High	Low
Enterprise fit	Native	Requires adapter
Scalability	Need capacity planning	Auto-scaling
Compliance	Full control	Twilio handles

Why would an enterprise prefer SIP integration over Twilio?

SIP is cheaper

Audio Codecs in SIP

Codec	Bitrate	Quality	Used In
G.711 (μ-law/a-law)	64 kbps	Fair	PSTN, default VoIP
G.722	64 kbps	Good	Wideband VoIP, enterprise
G.729	8 kbps	Fair	Low-bandwidth SIP
Opus	6-510 kbps	Excellent	Modern SIP, WebRTC gateway

SIP Integration with Voice Agent

# Using pjsua2 or baresip for SIP in Python
import pjsua2 as pj

class VoiceAgentSIPEndpoint(pj.Account):
    """SIP endpoint that connects calls to voice agent pipeline."""

    def onIncomingCall(self, call):
        """Handle incoming SIP call."""
        call_info = call.getInfo()
        print(f"Incoming call from: {call_info.remoteUri}")

        # Answer the call
        call.answer(200)

        # Get audio media
        aud_med = call_info.media[0]
        if aud_med.type == pj.MediaType.AUDIO:
            # Connect audio to voice agent pipeline
            self._connect_to_agent(aud_med)

    def _connect_to_agent(self, audio_media):
        """Connect SIP audio to ASR → LLM → TTS pipeline."""
        # Register audio callback
        audio_media.setRxCallback(self._on_audio_received)

    def _on_audio_received(self, frame):
        """Receive audio frame from SIP call."""
        # Feed to ASR
        asyncio.create_task(asr.process(frame))

SIP Server Options

Server	Language	Use Case
Asterisk	C	PBX, media server, voice agent gateway
FreeSWITCH	C	Scalable SIP, WebRTC gateway
Kamailio	C	SIP proxy/router (high scale)
OpenSIPS	C	SIP proxy, load balancer
pjsua2	C++/Python	SIP client library (embed in app)

FreeSWITCH as Voice Agent Gateway

FreeSWITCH is the most popular choice for connecting SIP to voice agents:

<!-- FreeSWITCH dialplan: route call to voice agent -->
<extension name="voice_agent">
  <condition field="destination_number" expression="^18005551234$">
    <action application="socket" data="127.0.0.1:8080 async full" />
  </condition>
</extension>

This connects the SIP call to your voice agent server via FreeSWITCH's ESL (Event Socket Layer), which streams audio bidirectionally.

Unlock the full lesson

You've read the first 2 sections. The rest of this lesson covers Build It, Use It, Ship It, Evaluation, Key Terms, Common Pitfalls, Interview Framing — plus a hands-on lab, quiz, and project artifact.

Create a free account to unlock Phase 0 and Phase 1 of every course — no credit card.

Key terms in this lesson

Agent