Create agent

Create a new agent with the provided configuration. The full config supports transcriber, model, voice, web search, transfer, end-call conditions, post-call actions (email + webhook), ambient background track, initial ringing sound, and multilingual support.

POST/agents/create

Bearersession

Body

19 fields

Name for the agent.

Initial message the agent will say when answering a call.

When true, the welcome message is treated as a directive the agent uses to generate a tailored greeting for each call, rather than being spoken word for word. When false, the welcome message is spoken exactly as written.

Allow the caller to interrupt the welcome message. When false, the agent finishes speaking the welcome before listening.

List of context breakdowns, each containing `title`, `body`, and optional `is_enabled`.

Title*

Title of the breakdown.

Body*

Body of the breakdown, the detailed prompt content.

Is enabled

Whether this section is included in the prompt.

curl -X POST 'https://backend.omnidim.io/api/v1/agents/create' \
  -H 'Authorization: Bearer YOUR_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
  "name": "Customer Support Agent",
  "welcome_message": "Hello! How can I help you today?",
  "dynamic_variables": {
    "customer_name": "Jane Doe",
    "order_id": "ORD-12345"
  },
  "context_breakdown": [
    {
      "title": "Purpose",
      "body": "This agent helps customers with product inquiries and support issues.",
      "is_enabled": true
    }
  ],
  "transcriber": {
    "provider": "deepgram_stream",
    "model": "nova-3",
    "language": "en-US",
    "silence_timeout_ms": 400,
    "interruption_min_words": 2,
    "max_call_duration_in_sec": 600
  },
  "model": {
    "model": "gpt-4.1-mini",
    "temperature": 0.7
  },
  "voice": {
    "provider": "eleven_labs",
    "voice_id": "JBFqnCBsd6RMkjVDRZzb",
    "model": "sonic-3.5",
    "speech_speed": 1
  },
  "web_search": {
    "provider": "DuckDuckGo"
  },
  "post_call_actions": {
    "email": {
      "recipients": [
        "support@example.com"
      ],
      "extracted_variables": [
        {
          "key": "customer_issue",
          "prompt": "Identify the main issue the customer is experiencing."
        }
      ],
      "trigger_call_statuses": [
        "completed",
        "voicemail_detected"
      ]
    },
    "webhook": {
      "url": "https://your-webhook-endpoint.com/omnidim-callback",
      "extracted_variables": [
        {
          "key": "customer_issue",
          "prompt": "Identify the main issue the customer is experiencing."
        }
      ],
      "trigger_call_statuses": [
        "completed",
        "failed"
      ]
    }
  },
  "transfer": {
    "transfer_options": [
      {
        "number": "+15551234567",
        "type": "static",
        "transfer_condition": "Transfer if the customer asks to speak with a human.",
        "transfer_message": "Please hold while I connect you to one of our agents."
      }
    ]
  },
  "end_call": {
    "condition": "End the call once the customer'\''s issue is resolved.",
    "message": "Thank you for contacting us. Have a great day!",
    "message_prompt": "End the call politely in the same language the user is speaking."
  },
  "background_track": {
    "volume": 0.2
  },
  "languages": [
    "English (India)",
    "Hindi"
  ]
}'

Example response

{
  "id": 6365,
  "name": "Customer Support Agent",
  "status": "Completed"
}

Authorization

BearerAuth

AuthorizationBearer <token>

Bearer token authentication. Obtain your API key from the OmniDimension dashboard.

In: header

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

name*string

Name for the agent.

welcome_message*string

Initial message the agent will say when answering a call.

is_welcome_message_dynamic?boolean

is_welcome_message_interruption?boolean

Allow the caller to interrupt the welcome message. When false, the agent finishes speaking the welcome before listening.

is_interruption_allowed?boolean

Global toggle for whether the caller can interrupt the agent mid-sentence at any point in the call.

dynamic_variables?

Key/value map used to substitute placeholders in the agent's prompt and welcome message at call time. Reference a variable in your prompt with {{variable_name}}. Useful for personalising the same agent across many calls.

context_breakdown*array<>

List of context breakdowns, each containing title, body, and optional is_enabled.

call_type?string

Call type of the assistant.

Value in"Incoming" | "Outgoing"

transcriber?

Configuration for the speech-to-text transcriber.

model?

Configuration for the language model.

voice?

Configuration for the text-to-speech voice.

web_search?

Configuration for web search capabilities.

post_call_actions?

Side effects that fire once the call ends. Configure email, webhook, or both.

transfer?

Conditional call transfer to a human agent or another number.

end_call?

Hang up automatically when a condition is met.

background_track?

Ambient background noise that plays under the agent's voice.

initial_ringing_sound_enabled?boolean

Plays a ringing tone after the call is picked up, until the agent starts speaking.

voicemail?

Voicemail / answering-machine handling for outbound calls. Set this with the nested object shown here; the agent object returns these values as the flat fields voicemail_enabled and voicemail_message. Voicemail detection is an access-gated feature. If it isn't enabled for your account, request access.

languages?array<string>

Languages the agent should support. Pass each language as a display-name string exactly as it appears in the dashboard's language picker. Unrecognized names are skipped.

Create agent

Authorization

Request Body

Response Body

`application/json`

Create agent

Authorization

Request Body

Response Body

200application/json

`application/json`