Custom API

Custom API integration connects your agent to external services through REST APIs, allowing it to fetch data or perform actions using third-party services.

Key benefits

• Access external data sources in real-time
• Connect to your own backend services
• Integrate with virtually any REST API endpoint
• Extend your agent's capabilities with custom functionality

Setup option 1: from the main dashboard

Setup option 2: from agent configuration

Setup option 3: from the chat interface

Setup option 4: with the API

Create the integration once, then attach it to any agent. Your API key goes in the Authorization: Bearer header. Resellers can do this for a client's agent by adding user_id. See Acting as clients.

curl -X POST "https://omnidim.io/api/v1/integrations/custom-api" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Order status",
    "description": "Look up an order'\''s delivery status by order number.",
    "url": "https://api.acme-corp.io/orders/status",
    "method": "POST",
    "headers": [{ "key": "Authorization", "value": "Bearer YOUR_TOKEN" }],
    "body_params": [
      {
        "key": "order_number",
        "description": "The order number the caller reads out",
        "type": "string",
        "required": true,
        "isLLMGenerated": true
      }
    ]
  }'

{
  "success": true,
  "integration_id": 4821,
  "integration": {
    "name": "Order status",
    "url": "https://api.acme-corp.io/orders/status",
    "method": "POST",
    "headers": [{ "key": "Authorization", "value": "********" }]
  }
}

curl -X POST "https://omnidim.io/api/v1/agents/158910/integrations" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "integration_id": 4821 }'

Configuration guide: basic settings

• Integration Name: choose a descriptive name (for example, "Weather API")
• Description: explain the API's purpose and usage. This is important for LLM tool calling.
• URL: full endpoint URL (for example, https://api.example.com/data)
• Method: select HTTP method (GET, POST, PUT, DELETE)

Configuration guide: headers

• Authorization tokens
• Content-Type specifications
• API keys
• Custom headers

Configuration guide: parameters

Specify both query parameters (GET) and body parameters (POST or PUT).

• Name: parameter identifier
• Description: purpose of this parameter
• Required: toggle between optional and required
• Data Type: select from dropdown
• AI Generated: toggle whether the AI should dynamically generate values

Testing your integration

Using with your agent

• Once connected, your agent will automatically use this API when appropriate
• Usage is determined based on the API description you provide

Passing the current call's phone numbers

With Pass call metadata on, every request your agent makes with POST, PUT, PATCH or DELETE gets these fields in its JSON body, when the call has them:

The switch is in the integration's Config tab. It is on by default for new integrations, including ones made with the API. Check it on older integrations before you rely on it. The Test tab sends sample values. A live call sends the real ones.

A GET request has no body, so these fields aren't sent. Use POST if your endpoint needs them. Turn the switch off if your endpoint rejects fields it doesn't expect.

The AI Generated setting is separate: it lets the agent fill a parameter from the conversation. Naming a parameter phone_number does not fill it with the call's number.

Example: Airtable

Example of using Airtable with custom API integration.

Example: Supabase user authentication

Walk through creating a complete Supabase authentication integration step by step.

Other example use cases

• CRM integration: retrieve customer data from Salesforce or HubSpot
• Inventory management: check product availability in real time
• User verification: authenticate users against internal databases
• Weather data: fetch current weather for location-based services
• Booking systems: check availability and make reservations

Pro tips

• Provide detailed descriptions for AI-generated parameters
• Use specific examples in your descriptions
• Test thoroughly before deploying to production