> ## Documentation Index
> Fetch the complete documentation index at: https://docs.eigi.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Create Agent
> Create a new AI agent with complete configuration
## Overview
Create a new AI agent with full configuration including speech-to-text (STT), speech-to-speech (STS), large language model (LLM), text-to-speech (TTS), main prompt, call analysis prompt, tools, and widget settings.
### Prompt Configuration
You can configure the agent's prompt in two ways:
1. **Inline prompt** (`prompt_content`) — Provide the prompt text directly. An internal prompt is auto-created for the agent.
2. **Prompt reference** (`prompt`) — Reference an existing prompt by name and optionally version.
### Call Analysis Configuration
You can optionally configure a dedicated call analysis prompt and schema:
1. **Enable analysis** (`call_analysis_enabled`) — Set this to `true` only when you want call analysis to run.
2. **Analysis prompt reference** (`call_analysis_prompt`) — Reference an existing analysis prompt by name and optionally version.
3. **Inline analysis prompt** (`call_analysis_prompt.custom_prompt`) — Provide prompt content directly to create a new analysis prompt version before linking it to the agent.
4. **Structured output schema** (`call_analysis_schema`) — Define the expected JSON structure for call analysis output.
If you do not want call analysis, omit `call_analysis_enabled`, `call_analysis_prompt`, and `call_analysis_schema`. If `call_analysis_enabled` is `true`, provide `call_analysis_prompt` as well.
### Try It Behavior
In `Try it`, `llm` is shown as a single object.
* For standard providers, fill `llm.params.temperature` only.
* For `CUSTOM_LLM`, fill `llm.params.base_url`, `llm.params.api_key`, and optionally `llm.params.API_TYPE`.
### Provider Restrictions (Public API)
| Provider | Allowed params |
| -------- | ------------------------------------------------------------------------------------------------------------------------------------------- |
| **STT** | No params allowed (must be empty `{}`) |
| **STS** | No public API param restrictions |
| **LLM** | Standard providers: only `temperature` (0.0 - 2.0). `CUSTOM_LLM`: requires `base_url` plus `api_key`; optional `API_TYPE` and `temperature` |
| **TTS** | Only `speed` |
## Authentication
Your eigi.ai API key. Must be prefixed with `vk_`.
## Request Body
Agent name (1-100 characters).
Agent type: `INBOUND` or `OUTBOUND`.
Speech-to-Text configuration.
STT provider: `DEEPGRAM`, `OPENAI`, `WHISPER`, or `GROK`.
Model name (e.g., `nova-2`).
Language code (e.g., `en`, `hi`).
Must be empty `{}` for public API requests.
Large Language Model configuration.
LLM provider: `OPENAI`, `OPENAI_RESPONSES`, `ANTHROPIC`, `GOOGLE`, `GROK`, `OPENROUTER`, or `CUSTOM_LLM`.
Model name (e.g., `gpt-4o`).
For standard LLM providers, only `temperature` (0.0-2.0) is allowed. For `CUSTOM_LLM`, provide `base_url`, `api_key`, and optionally `API_TYPE` (`CHAT_COMPLETION` or `RESPONSES`) and `temperature`. If `API_TYPE` is omitted, it defaults to `CHAT_COMPLETION`.
Text-to-Speech configuration.
TTS provider: `CARTESIA`, `DEEPGRAM`, `ELEVENLABS`, `SARVAM`, `GOOGLE`, `HUME`, or `GROK`.
Model name (e.g., `sonic-2`).
Language code.
Voice ID from the provider. Use the [List Voices](/api-reference/public/voices-list) endpoint to find available voices.
Only `speed` is allowed.
Speech-to-Speech configuration for realtime voice generation.
Whether speech-to-speech is enabled.
STS provider: `GOOGLE`, `OPENAI`, or `GROK`.
Model name for the STS provider.
Voice ID for the STS provider.
Language code.
Additional provider-specific STS parameters.
Agent description (max 500 characters).
Agent category (e.g., `Customer Service`, `Sales`, `Healthcare`, etc.).
Array of usage tip strings for the agent.
Mobile numbers to associate with the agent.
Inline prompt content. Auto-creates an internal prompt for the agent.
Reference to an existing prompt (alternative to `prompt_content`).
Name of the existing prompt.
Specific version (omit for latest).
Optional call analysis prompt configuration.
Name of an existing analysis prompt to reference.
Specific analysis prompt version to use (omit for latest).
Inline analysis prompt content. If provided, a new analysis prompt version is created before linking it to the agent.
Toggle call analysis on or off for the agent. Set this to `true` only when you
want call analysis to run. If `true`, provide `call_analysis_prompt` as well.
Optional schema describing the structured JSON output expected from call
analysis.
Agent's greeting message displayed when a chat session starts.
Configure agent tools. Each tool can be `true` (enable), `false` (disable), or an object with configuration.
Available tools: `end_call`, `transfer_call`, `schedule_call`, `set_language`, `background_audio`, `user_idle`.
Widget configuration for chat interface customization.
## Response
Returns the complete agent object with all configurations, `id`, and timestamps.
## OpenAPI
````yaml POST /v1/public/agents
openapi: 3.0.0
info:
title: eigi.ai API
description: REST API for managing AI voice agents and conversations
version: 1.0.0
contact:
email: buddy@eigi.ai
url: https://eigi.ai
servers:
- url: https://api.eigi.ai
description: Production server
security:
- ApiKeyAuth: []
paths:
/v1/public/agents:
post:
tags:
- Agents
summary: Create Agent
description: >-
Create a new AI agent with complete configuration including STT, LLM,
TTS, prompt, call analysis configuration, tools, and widget settings.
Supports two modes for prompt: inline prompt_content (auto-creates
internal prompt) or reference to an existing prompt by name.
operationId: createAgent
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/AgentCreate'
examples:
basic:
summary: Create agent with inline prompt
value:
agent_name: Support Agent
agent_type: INBOUND
agent_category: Customer Service
stt:
provider_name: DEEPGRAM
model_name: nova-2
language: en
params: {}
llm:
provider_name: OPENAI
model_name: gpt-4o
params:
temperature: 0.7
tts:
provider_name: CARTESIA
model_name: sonic-2
language: en
voice_id: a0e99841-438c-4a64-b679-ae501e7d6091
params:
speed: 1
prompt_content: You are a helpful customer support agent.
call_analysis_enabled: true
call_analysis_prompt:
prompt_name: support_call_analysis
custom_prompt: >-
Analyze the call outcome, customer sentiment, and
follow-up actions.
call_analysis_schema:
call_outcome:
type: str
customer_sentiment:
type: str
follow_up_required:
type: bool
first_message_prompt: Hello! How can I help you today?
tools:
end_call: true
transfer_call:
transfer_phone_number: '+919876543210'
custom_llm:
summary: Create agent with CUSTOM_LLM
value:
agent_name: Custom LLM Support Agent
agent_type: OUTBOUND
agent_category: Customer Service
stt:
provider_name: DEEPGRAM
model_name: nova-2
language: en
params: {}
llm:
provider_name: CUSTOM_LLM
model_name: custom-model
params:
base_url: https://llm.example.com/v1
api_key: sk-...
API_TYPE: RESPONSES
tts:
provider_name: CARTESIA
model_name: sonic-2
language: en
voice_id: a0e99841-438c-4a64-b679-ae501e7d6091
params:
speed: 1
prompt_content: You are a helpful support agent using a custom LLM backend.
first_message_prompt: Hello! How can I help you today?
responses:
'201':
description: Agent created successfully
content:
application/json:
schema:
$ref: '#/components/schemas/AgentResponse'
'400':
description: Bad request - Missing or invalid parameters
content:
application/json:
schema:
$ref: '#/components/schemas/Error'
'401':
description: Unauthorized - Invalid API key
content:
application/json:
schema:
$ref: '#/components/schemas/Error'
'404':
description: Not found
content:
application/json:
schema:
$ref: '#/components/schemas/Error'
'500':
description: Internal server error
content:
application/json:
schema:
$ref: '#/components/schemas/Error'
security:
- ApiKeyAuth: []
components:
schemas:
AgentCreate:
type: object
required:
- agent_name
- agent_type
- stt
- llm
- tts
properties:
agent_name:
type: string
minLength: 1
maxLength: 100
description: Agent name (1-100 characters)
example: Customer Support Agent
agent_description:
type: string
maxLength: 500
description: Agent description (max 500 characters)
agent_type:
type: string
enum:
- INBOUND
- OUTBOUND
description: Agent type
agent_category:
type: string
enum:
- Customer Service
- Sales
- Scheduling
- Research
- Healthcare
- Customer Success
- Education
- Finance
- HR
- Legal
- Productivity
- Creative
- IT/DevOps
- E-commerce
- Entertainment
- Travel
- Wellness
- Anime
- Cartoon
- Gaming
- Lifestyle
- other
default: other
description: Agent category
agent_tips:
type: array
items:
type: string
description: Usage tips for the agent
agent_mobile_number:
type: array
items:
type: string
description: Mobile numbers associated with the agent
stt:
$ref: '#/components/schemas/STTConfig'
llm:
$ref: '#/components/schemas/LLMConfig'
tts:
$ref: '#/components/schemas/TTSConfig'
call_analysis_prompt:
type: object
description: >-
Optional call analysis prompt reference. Provide custom_prompt to
create a new analysis prompt version before linking it to the agent.
properties:
prompt_name:
type: string
description: Name of existing analysis prompt
prompt_version:
type: integer
description: Specific version (omit for latest)
custom_prompt:
type: string
description: Inline analysis prompt content
required:
- prompt_name
call_analysis_enabled:
type: boolean
description: >-
Toggle call analysis on or off for the agent. Set to true only when
you want call analysis to run; when true, provide
call_analysis_prompt as well.
call_analysis_schema:
type: object
description: >-
Optional schema for structured call analysis output, keyed by field
name.
additionalProperties:
type: object
properties:
type:
type: string
enum:
- str
- int
- float
- bool
- list
- dict
- Optional[str]
- Optional[int]
- Optional[float]
- Optional[bool]
- Optional[list]
- Optional[dict]
description: Expected type for this call analysis field.
required:
- type
additionalProperties: false
example:
summary:
type: str
customer_intent:
type: Optional[str]
follow_up_required:
type: bool
prompt_content:
type: string
description: >-
Inline prompt content — auto-creates an internal prompt for the
agent
prompt:
type: object
description: Reference to an existing prompt (alternative to prompt_content)
properties:
prompt_name:
type: string
description: Name of existing prompt
prompt_version:
type: integer
description: Specific version (omit for latest)
required:
- prompt_name
first_message_prompt:
type: string
description: Agent's greeting message shown when chat starts
tools:
$ref: '#/components/schemas/ToolsConfig'
widget_config:
type: object
description: Widget configuration for chat interface customization
properties:
widget_interface:
type: object
description: Interface settings (enable_chat, mute_button, etc.)
widget_appearance:
type: object
description: Appearance settings (variant, placement, etc.)
widget_avatar:
type: object
description: Avatar settings (type, colors)
widget_tnc_config:
type: object
description: Terms and conditions settings
AgentResponse:
type: object
properties:
id:
type: string
description: Agent ID
agent_name:
type: string
description: Agent name
agent_description:
type: string
description: Agent description
agent_type:
type: string
enum:
- INBOUND
- OUTBOUND
description: Agent type
agent_category:
type: string
description: Agent category
agent_tips:
type: array
items:
type: string
description: Usage tips
agent_mobile_number:
type: array
items:
type: string
description: Associated mobile numbers
stt:
type: object
description: STT configuration
llm:
type: object
description: LLM configuration
tts:
type: object
description: TTS configuration
sts:
type: object
description: STS configuration (if set)
call_analysis_enabled:
type: boolean
description: Whether call analysis is enabled for the agent
prompt:
type: object
description: Prompt reference
properties:
prompt_name:
type: string
prompt_version:
type: integer
prompt_content:
type: string
first_message_prompt:
type: string
description: Agent's greeting message
dynamic_variables:
type: array
items:
type: string
description: Dynamic variables in the prompt
tools:
type: object
description: Configured tools
widget_config:
type: object
description: Widget configuration
MediaFiles:
type: array
items:
type: object
properties:
fileType:
type: string
mediaUrl:
type: string
description: Media files associated with the agent
created_at:
type: string
format: date-time
description: Creation timestamp
updated_at:
type: string
format: date-time
description: Last update timestamp
Error:
type: object
properties:
detail:
type: string
description: Error message
STTConfig:
type: object
required:
- provider_name
- model_name
- language
properties:
provider_name:
type: string
enum:
- DEEPGRAM
- WHISPER
description: STT provider name
model_name:
type: string
description: Model name for the STT provider
example: nova-2
language:
type: string
description: Language code (e.g., en, hi, es, fr)
example: en
params:
type: object
description: Must be empty for public API - no additional STT params allowed
example: {}
LLMConfig:
type: object
required:
- provider_name
- model_name
description: >-
LLM configuration. Use provider_name to choose a standard provider or
CUSTOM_LLM. For standard providers, only params.temperature applies. For
CUSTOM_LLM, use params.base_url, params.api_key, and optional
params.API_TYPE.
properties:
provider_name:
type: string
enum:
- OPENAI
- OPENAI_RESPONSES
- ANTHROPIC
- GOOGLE
- GROK
- OPENROUTER
- CUSTOM_LLM
description: LLM provider name
model_name:
type: string
description: Model name for the selected LLM provider
example: gpt-4o
params:
type: object
description: >-
For standard providers, only temperature is used. For CUSTOM_LLM,
provide base_url and api_key, and optionally API_TYPE.
properties:
temperature:
type: number
minimum: 0
maximum: 2
description: Standard providers only. Controls randomness (0.0-2.0).
base_url:
type: string
description: CUSTOM_LLM only. Base URL for the custom LLM endpoint.
example: https://llm.example.com/v1
api_key:
type: string
description: CUSTOM_LLM only. API key for the custom LLM provider.
example: sk-...
API_TYPE:
type: string
enum:
- CHAT_COMPLETION
- RESPONSES
description: >-
CUSTOM_LLM only. Optional; defaults to CHAT_COMPLETION when
omitted.
example:
temperature: 0.7
TTSConfig:
type: object
required:
- provider_name
- model_name
- language
- voice_id
properties:
provider_name:
type: string
enum:
- CARTESIA
- ELEVENLABS
- SARVAM
- GOOGLE
- HUME
description: TTS provider name
model_name:
type: string
description: Model name for the TTS provider
example: sonic-2
language:
type: string
description: Language code (e.g., en, hi)
example: en
voice_id:
type: string
description: Voice ID from the TTS provider
example: a0e99841-438c-4a64-b679-ae501e7d6091
params:
type: object
description: Only 'speed' param is allowed for public API
properties:
speed:
type: number
description: Speech speed multiplier
example:
speed: 1
ToolsConfig:
type: object
description: >-
Configure agent tools. Each tool can be true (enable), false (disable),
or an object with configuration.
properties:
end_call:
description: End call tool - enabled by default. Use true/false to toggle.
oneOf:
- type: boolean
- type: object
transfer_call:
description: >-
Transfer call tool. When using config object, requires
transfer_phone_number.
oneOf:
- type: boolean
- type: object
properties:
transfer_phone_number:
type: string
description: Phone number to transfer to
required:
- transfer_phone_number
schedule_call:
description: Schedule call tool. Optional allow_immediate config.
oneOf:
- type: boolean
- type: object
properties:
allow_immediate:
type: boolean
default: false
set_language:
description: Set language tool. Optional language_string config.
oneOf:
- type: boolean
- type: object
properties:
language_string:
type: string
default: en-US
background_audio:
description: Background audio tool with audio file and volume config.
oneOf:
- type: boolean
- type: object
properties:
audio_files:
type: array
items:
type: string
enum:
- env1.wav
- honk1.wav
- market1.wav
- office1.mp3
- office2.mp3
- office3.mp3
- office4.wav
- office5.wav
- temple1.wav
- temple2.wav
background_noise_reduce:
type: boolean
default: true
default_volume:
type: number
minimum: 0.1
maximum: 1
default: 0.2
user_idle:
description: User idle detection with timeout and retry config.
oneOf:
- type: boolean
- type: object
properties:
idle_timeout_seconds:
type: number
minimum: 3
maximum: 30
default: 10
max_retry_count:
type: integer
minimum: 1
maximum: 10
default: 3
securitySchemes:
ApiKeyAuth:
type: apiKey
in: header
name: X-API-Key
description: >-
API key for authentication. Get your API key from the eigi.ai Dashboard
under Settings → API Keys.
````