Go to app.chanl.ai/signup and create a free account. You'll land on the dashboard.

Navigate to Build → MCP in the sidebar. Choose your platform and follow the setup wizard.

# In your VAPI dashboard, add Chanl as an MCP server
# Server URL: https://mcp.chanl.ai
# API Key: your Chanl API key from Settings → API Keys

# In your Retell dashboard, configure the MCP endpoint
# Endpoint: https://mcp.chanl.ai
# Authentication: Bearer token from Chanl Settings

curl -X POST "https://api.chanl.ai/api/v1/agents/connect" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "provider": "custom",
    "name": "My Agent",
    "phoneNumber": "+1234567890"
  }'

Once connected, your agent appears in the Agents list on the dashboard.

Personas are the AI-generated callers that test your agent. Go to Test → Personas and create a few:

Test Personas

Simulate real caller behaviors

6 active

Angry Caller

Frustrated

Polite Pro

Professional

Confused User

Uncertain

VIP Client

High-value

Tech Savvy

Expert

First-Time Caller

Nervous

Or create them via API:

curl -X POST "https://api.chanl.ai/api/v1/personas" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Frustrated Customer",
    "description": "Impatient caller who has been on hold too long",
    "emotionalTone": "frustrated",
    "speakingStyle": "interrupting"
  }'

A scenario combines your agent + personas + a scorecard into an automated test.

Go to Test → Scenarios → Create Scenario, then:

1. Name it (e.g., "Refund Request Handling")
2. Describe the situation the caller is in
3. Select your personas
4. Select your agent
5. Choose a scorecard (use the default "Customer Service" scorecard to start)
6. Click Run

Chanl will run a realistic phone conversation between each persona and your agent, then score the results.

After simulations complete (usually 1-2 minutes), go to Observe → Analytics to see:

• Overall score — how your agent performed across all test calls
• Per-persona breakdown — which customer types your agent handles well vs. poorly
• Full transcripts — read every word of every test conversation
• Scorecard results — detailed pass/fail on each evaluation criterion