OpenClaw Voice & SMS Integration (v3.0)

Fully integrated voice and SMS system that connects directly to your OpenClaw session.

What's New in v3.0

• ✅ Real OpenClaw Integration: Speech/SMS → OpenClaw session → Intelligent response
• ✅ Full Context: Access to weather, calendar, memory, skills, everything
• ✅ Session Memory: Maintains conversation context across calls
• ✅ Dynamic Responses: No hardcoded responses, real AI thinking
• ✅ ElevenLabs Premium Voice: Natural-sounding speech with Polly fallback

How It Works

User calls/texts → Twilio → Webhook → OpenClaw API → Response → TTS → User

Every interaction goes through OpenClaw's full AI pipeline with access to:

• Weather service
• Calendar integrations
• Browser automation results
• Memory/context
• All installed skills

Environment Variables

Required

• GATEWAY_URL - Your OpenClaw Gateway URL (e.g., https://gateway.openclaw.ai)
• GATEWAY_TOKEN - OpenClaw API authentication token
• TWILIO_ACCOUNT_SID - Twilio Account SID
• TWILIO_AUTH_TOKEN - Twilio Auth Token

Optional (for premium voice)

• ELEVENLABS_API_KEY - ElevenLabs API key
• ELEVENLABS_VOICE_ID - Voice ID (default: Adam)
• AWS_ACCESS_KEY_ID - AWS credentials for Polly fallback
• AWS_SECRET_ACCESS_KEY - AWS credentials for Polly fallback

Deployment

Render.com

1. Create Web Service from this repo
2. Add Environment Variables (see above)
3. Set Start Command: npm start
4. Upgrade to Paid Tier ($7/mo) - eliminates cold starts

Railway / Fly.io

Similar process - add environment variables and deploy.

Twilio Configuration

1. Go to Twilio Console → Phone Numbers
2. Set Voice Webhook: https://your-service.onrender.com/voice
3. Set SMS Webhook: https://your-service.onrender.com/sms
4. Save changes

Testing

Call or text your Twilio number. The system will:

1. Receive your speech/text
2. Send it to OpenClaw for processing
3. Get intelligent response with full context
4. Convert to speech (for calls) or text (for SMS)
5. Play/send response back to you

Local Development

npm install
export GATEWAY_URL=https://your-gateway.openclaw.ai
export GATEWAY_TOKEN=your_token_here
export TWILIO_ACCOUNT_SID=ACxxxxx
export TWILIO_AUTH_TOKEN=xxxxx
npm start

OpenClaw Gateway Token

To get your Gateway token:

1. Log into OpenClaw web interface
2. Go to Settings → API
3. Generate a new token with sessions.send permission
4. Copy token and add to environment variables

Architecture

• Express.js - Web server
• Twilio SDK - TwiML generation
• Axios - OpenClaw API calls
• ElevenLabs - Premium TTS (primary)
• AWS Polly - TTS fallback
• Audio Caching - Fast responses for repeated phrases

Pricing Considerations

• Render Free Tier: Has cold starts (30s delay)
• Render Paid ($7/mo): Always-on, instant response
• ElevenLabs: ~$0.30 per 1000 characters
• AWS Polly: ~$4 per 1 million characters
• Twilio: ~$0.01-0.02 per minute for calls

Voice Quality

ElevenLabs (when configured):

• Natural, human-like speech
• Multiple voice options
• Emotional range
• Premium pricing

AWS Polly (fallback):

• Neural voices available
• Reliable, fast
• Lower cost
• Slightly robotic

Session Management

Each phone number gets its own OpenClaw session with label voice-{phone_number}.

This means:

• Conversation context is maintained
• Memory persists across calls
• Can reference previous interactions
• "Remember when I asked about weather earlier?"

Troubleshooting

Application Error on call:

• Service is sleeping (free tier) - call again or upgrade
• Check Render logs for errors

"I'm having trouble connecting":

• Check GATEWAY_URL and GATEWAY_TOKEN are correct
• Verify OpenClaw Gateway is running
• Check Render logs for API errors

Robotic voice:

• ElevenLabs not configured, using Polly fallback
• Add ELEVENLABS_API_KEY and ELEVENLABS_VOICE_ID

No response:

• OpenClaw session timeout (increase timeoutSeconds in code)
• Check OpenClaw Gateway logs

License

MIT