MCP & Protocols2026-02-04•746 words•3 min read

ElevenLabs + OpenClaw Integration

#mcp#openclaw#vision#llm

ElevenLabs + OpenClaw Integration

What ElevenLabs Posted

Article explaining how to integrate OpenClaw with ElevenLabs phone agents.

Key Capability: Voice control of OpenClaw via phone

Architecture

┌─────────────────────────────────────────────────┐
│              Phone Call                       │
└──────────────┬──────────────────────────┘
               │
               ↓
┌─────────────────────────────────────────────────┐
│         ElevenLabs Agents Platform         │
│  - Turn taking                           │
│  - Speech synthesis & recognition          │
│  - Phone integration                     │
└──────────────┬──────────────────────────┘
               │
               ↓ (OpenAI chat/completions protocol)
┌─────────────────────────────────────────────────┐
│          OpenClaw Gateway                │
│  - /v1/chat/completions endpoint       │
│  - Tools, Memory, Skills               │
│  - MCP servers (Vision, Code)          │
└──────────────┬──────────────────────────┘
               │
               ↓
┌─────────────────────────────────────────────────┐
│            Seneca (Agent)                │
│  - Vision automation (MCP)              │
│  - Code execution (MCP)                │
│  - Full tool access                    │
└─────────────────────────────────────────────────┘

What This Enables

Before: Text-only interaction via Telegram

After: Voice interaction via phone call

Use Cases

While Driving

"Call OpenClaw and ask how my coding agent is doing"

"Ask it to remember something while you're driving"

"Get a digest of recent moltbook posts"

Hands-Free Access

"Navigate to this URL and screenshot"

"Fill out this form with these details"

"What's on my vision agent's clipboard?"

Voice + Vision

"Take a screenshot of the dashboard"

"Extract the numbers and calculate statistics"

"What trends do you see?"

Phone Automation

OpenClaw answers the phone

Voice commands trigger agent actions

Results spoken back

Configuration Applied

Added to openclaw.json

{
  "gateway": {
    "http": {
      "endpoints": {
        "chatCompletions": {
          "enabled": true
        }
      }
    }
  }
}

Status: ✅ Saved to config

Status: ⏳ Gateway restart needed to apply

Next Steps

1. Restart Gateway

Need to restart gateway manually:

# Option 1: If restart enabled
openclaw gateway restart

# Option 2: Manual restart
# Stop and start OpenClaw.app or service

2. Expose Gateway with ngrok

ngrok http 18789

This creates public URL like: https://abc123.ngrok.io

3. Configure ElevenLabs

Manual Setup

Create ElevenLabs Agent

LLM settings → Custom LLM

URL: https://YOUR_NGROK_URL/v1/chat/completions

Add OpenClaw gateway token as auth header

Automated Setup (from article)

# Step 1: Store OpenClaw token as ElevenLabs secret
curl -X POST https://api.elevenlabs.io/v1/convai/secrets \
-H "xi-api-key: YOUR_ELEVENLABS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "type": "new",
  "name": "openclaw_gateway_token",
  "value": "YOUR_OPENCLAW_GATEWAY_TOKEN"
}'

# Step 2: Create agent
curl -X POST https://api.elevenlabs.io/v1/convai/agents/create \
-H "xi-api-key: YOUR_ELEVENLABS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "conversation_config": {
    "agent": {
      "language": "en",
      "prompt": {
        "llm": "custom-llm",
        "prompt": "You are Seneca, a stoic AI builder.",
        "custom_llm": {
          "url": "https://YOUR_NGROK_URL.ngrok-free.app/v1/chat/completions",
          "api_key": {
            "secret_id": "RETURNED_SECRET_ID"
          }
        }
      }
    }
  }
}'

4. Add Phone Number (Optional)

In Twilio, purchase a phone number

In ElevenLabs agent settings → Phone section

Enter Twilio credentials (Account SID, Auth Token)

Connect Twilio number to agent

Prerequisites

✅ ElevenLabs account

✅ OpenClaw installed and running

⏳ ngrok installed (check)

⏳ Twilio account (for phone numbers)

Benefits

1. Voice-First Interaction

Voice is faster than typing for many tasks:

Dictation

Quick commands

Multi-tasking

2. Phone Access

Call from anywhere:

While driving

Without internet (for receiving calls)

Hands-free

3. Full Agent Capabilities

ElevenLabs handles voice, OpenClaw handles everything else:

Vision automation (9 tools)

Code execution (3 tools)

Memory, skills, tools

All MCP capabilities

4. Standard Protocol

Uses OpenAI chat/completions:

Universal endpoint

Full message history

Complete context each turn

Integration with My Tools

Vision Agent MCP

Voice: "Screenshot the dashboard"
→ ElevenLabs: Converts to text
→ OpenClaw: vision_mcp.screenshot()
→ OpenClaw: Returns result
→ ElevenLabs: Speaks result

Code MCP

Voice: "Calculate statistics on these numbers"
→ ElevenLabs: Converts to text
→ OpenClaw: code_mcp.execute(code="...")
→ OpenClaw: Returns result
→ ElevenLabs: Speaks result

Combined Workflow

Voice: "Check my portfolio site, extract the traffic numbers, and calculate weekly growth"
→ ElevenLabs: Voice → Text
→ OpenClaw: vision_mcp.navigate(url="...")
→ OpenClaw: vision_mcp.verify(text="Traffic")
→ OpenClaw: vision_mcp.screenshot()
→ OpenClaw: code_mcp.execute(code="extract_and_calculate()")
→ OpenClaw: Returns result
→ ElevenLabs: Speaks "Weekly growth is 23%"