8bit-ai

Phone / SIP Integration

Add voice capabilities to your AI agent and handle phone calls the same way you handle chat conversations. Connect via SIP trunking, Twilio, or Vonage to enable natural voice conversations with text-to-speech and speech-to-text pipelines.

app.8bit-ai.com
Phone/SIP integration

Speech-to-Text

Automatic transcription of caller speech with support for multiple languages

Text-to-Speech

Natural-sounding voice responses with configurable speed, pitch, and accent

Multiple Providers

Connect via SIP trunking, Twilio, or Vonage for maximum flexibility

Overview

The phone integration turns your AI agent into a voice-enabled assistant that can answer calls, conduct conversations, and interact with callers using natural speech. When a call comes in, the audio is streamed through a speech-to-text engine, the transcribed text is sent to your agent, and the agent's response is converted back to speech via text-to-speech. The entire pipeline operates in real-time with sub-second latency.

Call Flow

  1. Incoming call arrives via SIP trunk or Twilio/Vonage
  2. Call is routed to 8bit-ai's media server
  3. Caller's speech is transcribed to text via STT engine
  4. Transcribed text is sent to your AI agent for processing
  5. Agent generates a response using its LLM and knowledge base
  6. Response is converted to speech via TTS engine
  7. Audio is streamed back to the caller in real-time

Supported Providers

SIP Trunking

Direct SIP connection to your PBX or SIP provider

Twilio

Twilio Phone Numbers + Programmable Voice API

Vonage (Nexmo)

Vonage Voice API with SIP and WebSocket support

Voice Pipeline

The phone integration requires the voice pipeline to be enabled in your agent settings. You can configure separate TTS and STT models, adjust voice parameters like speed and pitch, and select language preferences. Voice pipeline is available on Pro and Enterprise plans.

SIP Trunking

Connect your AI agent directly via SIP trunking for integration with existing PBX systems or enterprise telephony infrastructure.

1

Configure SIP Endpoint

Set up the SIP trunk in your network that will route calls to 8bit-ai.

SIP Server

sip.8bit-ai.com:5060

Transport

UDP / TCP / TLS

Codecs

PCMA, PCMU, G.722, Opus

Authentication

IP-based (whitelist your IP range)
2

Add SIP Integration

Configure the SIP connection in the 8bit-ai dashboard.

  1. aGo to Integrations in your dashboard
  2. bClick Phone / SIP and select SIP Trunk
  3. cEnter a name for your integration
  4. dConfigure the SIP URI or IP whitelist for inbound calls
  5. eSelect your voice pipeline preferences (TTS/STT models)
  6. fClick Connect to activate the trunk
3

Link to Agent

Connect the SIP trunk to your AI agent. Once linked, incoming calls on the trunk will be answered by the agent.

  1. aNavigate to your agent's Integrations tab
  2. bFind your SIP integration and click Add to Agent
  3. cConfigure call settings (greeting, max duration, language)
  4. dTest by placing a call to the SIP trunk

Network Requirements

SIP trunking requires your network to allow SIP traffic on port 5060 (UDP/TCP) and RTP media traffic on ports 10000-20000 (UDP). Ensure your firewall is configured accordingly. For TLS connections, port 5061 must be open.

Twilio Setup

Connect your AI agent to existing Twilio phone numbers with minimal configuration. The integration uses Twilio's Programmable Voice API with WebSocket streaming for real-time audio processing.

1

Create Twilio Integration

Set up the connection between 8bit-ai and your Twilio account.

  1. aGo to Integrations > Phone / SIP > Twilio
  2. bEnter a name for your integration
  3. cPaste your Twilio Account SID and Auth Token
  4. dSelect your voice pipeline preferences
  5. eClick Connect to link your Twilio account
Twilio Voice API documentation
2

Configure Twilio Phone Number

Point your Twilio phone number to the 8bit-ai webhook for incoming calls.

  1. aLog in to your Twilio Console
  2. bGo to Phone Numbers > Manage > Active Numbers
  3. cSelect the number you want to connect
  4. dSet Voice > When a call comes in to Webhook
  5. eEnter the webhook URL provided in the 8bit-ai dashboard
  6. fSet method to POST and save

Webhook URL format:

https://your-domain.com/api/twilio/voice/{agent_id}
3

Link to Agent

Connect the Twilio integration to your agent.

  1. aGo to your agent's Integrations tab
  2. bFind your Twilio integration and click Add to Agent
  3. cConfigure call settings (greeting, timeout, language)
  4. dCall your Twilio number to test the integration

Phone Numbers

Manage which phone numbers are connected to your agent and configure number-specific behaviors.

Adding Phone Numbers

  • Purchase numbers through Twilio, Vonage, or your SIP provider
  • Register the number in the 8bit-ai Phone Integration settings
  • Assign the number to an agent via the Integrations tab
  • Verify by placing a test call to the number

Number Configuration

  • Set a custom greeting message per number
  • Configure language and locale per number
  • Define max call duration (default: 30 minutes)
  • Set voice model preferences (TTS voice, STT language)

Routing Rules

Configure how calls are routed based on the number dialed:

Porting Existing Numbers

You can port existing phone numbers to Twilio or Vonage. The porting process typically takes 5-10 business days. During porting, configure your current provider to forward calls to the 8bit-ai SIP endpoint.

Twilio number porting guide

Call Handling

Configure how your agent handles incoming calls, including greetings, interruptions, fallback behaviors, and call termination.

Inbound Call Behavior

Greeting

When a call is answered, the agent plays a greeting message before listening. Configure this in the integration or agent settings. Example: "Hello, you've reached Acme support. How can I help you today?"

Speech Detection

The system uses Voice Activity Detection (VAD) to determine when the caller has finished speaking. Adjust the silence threshold and timeout in the voice pipeline settings.

Interruption Handling

Callers can interrupt the agent mid-response. The agent will stop speaking, listen to the new input, and incorporate the interruption into its context.

Call Termination

The agent can end the call when the conversation is complete, or the caller can hang up at any time. Configure a closing message before disconnection.

Call Configuration Options

Outbound Calls

The phone integration also supports outbound calls initiated by your agent or through the API.

  1. Trigger an outbound call via the API or agent action
  2. Specify the destination number and caller ID
  3. The system places the call and connects the agent when answered
  4. The agent delivers its message and listens for the recipient's response

Regulatory Compliance

When making outbound calls, ensure compliance with local telemarketing and communications regulations. This includes maintaining Do-Not-Call lists, providing opt-out mechanisms, and adhering to calling hours. Consult your legal team for jurisdiction-specific requirements.

Troubleshooting

No Audio / One-Way Audio

Possible causes:

  • Firewall blocking RTP media ports (10000-20000 UDP)
  • NAT traversal issues with SIP signaling
  • Incorrect codec configuration

Solutions:

  • Ensure UDP ports 10000-20000 are open for RTP traffic
  • Verify SIP ALG is disabled on your firewall
  • Confirm codecs match between your PBX and 8bit-ai
  • Test with a different transport protocol (UDP/TCP/TLS)

Call Drops Immediately

Possible causes:

  • Agent not in Active status
  • Integration not linked to an agent
  • SIP registration failed

Solutions:

  • Verify agent status is "Active"
  • Confirm the integration is linked in the agent's Integrations tab
  • Check SIP registration status in the integration settings

Poor Transcription Accuracy

Possible causes:

  • Background noise on the caller's end
  • Incorrect language/locale configuration
  • Low audio quality from the phone network

Solutions:

  • Enable noise reduction in voice pipeline settings
  • Verify the language setting matches the caller's language
  • Try a different STT model (e.g., switch to Whisper)
  • Use a higher-quality audio codec (Opus or G.722)

Call Logs

Review detailed call logs in the Sessions dashboard. Each call session includes the full transcript, duration, audio quality metrics, and any errors encountered during the call. Use these logs to debug issues and optimize your configuration.

Related Documentation

Integrations Overview

Compare all available integration channels

Session Monitoring

Track phone conversations and transcripts

AI Agents

Configure voice pipeline and agent settings