Create Agent

curl -X POST https://api.thinnest.ai/v2/agents \
  -H "Authorization: Bearer $THINNESTAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "name": "Sales Agent",
  "model": "gpt-4o",
  "instructions": "You are a helpful sales assistant.",
  "description": "Handles product inquiries and purchase decisions",
  "voiceEnabled": true,
  "tools": [
    "duckduckgo",
    "email"
  ],
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  },
  "voice": {
    "provider": "deepgram",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  },
  "firstMessage": "Hello! How can I help you today?",
  "temperature": 0.7,
  "maxTokens": 4096
}'

import requests

url = "https://api.thinnest.ai/v2/agents"
headers = {
    "Authorization": "Bearer " + "YOUR_THINNESTAI_API_KEY",
    "Content-Type": "application/json",
}

payload = {
  "name": "Sales Agent",
  "model": "gpt-4o",
  "instructions": "You are a helpful sales assistant.",
  "description": "Handles product inquiries and purchase decisions",
  "voiceEnabled": true,
  "tools": [
    "duckduckgo",
    "email"
  ],
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  },
  "voice": {
    "provider": "deepgram",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  },
  "firstMessage": "Hello! How can I help you today?",
  "temperature": 0.7,
  "maxTokens": 4096
}

response = requests.post(url, headers=headers, json=payload)
print(response.json())

const response = await fetch("https://api.thinnest.ai/v2/agents", {
  method: "POST",
  headers: {
    "Authorization": "Bearer " + "YOUR_THINNESTAI_API_KEY",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
  "name": "Sales Agent",
  "model": "gpt-4o",
  "instructions": "You are a helpful sales assistant.",
  "description": "Handles product inquiries and purchase decisions",
  "voiceEnabled": true,
  "tools": [
    "duckduckgo",
    "email"
  ],
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  },
  "voice": {
    "provider": "deepgram",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  },
  "firstMessage": "Hello! How can I help you today?",
  "temperature": 0.7,
  "maxTokens": 4096
}),
});

const data = await response.json();
console.log(data);

package main

import (
    "bytes"
    "fmt"
    "io"
    "net/http"
)

func main() {
    payload := []byte(`{
  "name": "Sales Agent",
  "model": "gpt-4o",
  "instructions": "You are a helpful sales assistant.",
  "description": "Handles product inquiries and purchase decisions",
  "voiceEnabled": true,
  "tools": [
    "duckduckgo",
    "email"
  ],
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  },
  "voice": {
    "provider": "deepgram",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  },
  "firstMessage": "Hello! How can I help you today?",
  "temperature": 0.7,
  "maxTokens": 4096
}`)
    req, _ := http.NewRequest("POST", "https://api.thinnest.ai/v2/agents", bytes.NewBuffer(payload))
    req.Header.Set("Authorization", "Bearer YOUR_THINNESTAI_API_KEY")
    req.Header.Set("Content-Type", "application/json")

    resp, err := http.DefaultClient.Do(req)
    if err != nil { panic(err) }
    defer resp.Body.Close()
    body, _ := io.ReadAll(resp.Body)
    fmt.Println(string(body))
}

{
  "id": "ag_c47e7c97_b2f2",
  "name": "Sales Support Agent",
  "createdAt": "2026-03-07T10:00:00Z",
  "updatedAt": null,
  "model": {
    "provider": "openai",
    "model": "gpt-4o",
    "temperature": 0.7,
    "maxTokens": 4096,
    "instructions": "You are a sales support agent for Acme Corp..."
  },
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  },
  "voice": {
    "provider": "deepgram",
    "model": "aura-2",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  },
  "tools": [
    { "name": "duckduckgo", "type": "duckduckgo", "enabled": true },
    { "name": "email", "type": "email", "enabled": true },
    { "name": "calculator", "type": "calculator", "enabled": true },
    { "name": "googlesheets", "type": "googlesheets", "enabled": true }
  ],
  "knowledgeBase": {
    "sources": [
      { "id": "1", "name": "Product Catalog", "type": "file", "status": "ready" },
      { "id": "3", "name": "Pricing FAQ", "type": "text", "status": "ready" }
    ]
  },
  "voiceEnabled": true,
  "agentType": "simple",
  "firstMessage": "Hi there! Thanks for calling Acme Corp. How can I help you today?",
  "endCallEnabled": true,
  "maxDurationSeconds": 900,
  "noiseCancellation": "krisp"
}

curl -X POST https://api.thinnest.ai/v2/agents \
  -H "Authorization: Bearer $THINNESTAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "name": "Sales Agent",
  "model": "gpt-4o",
  "instructions": "You are a helpful sales assistant.",
  "description": "Handles product inquiries and purchase decisions",
  "voiceEnabled": true,
  "tools": [
    "duckduckgo",
    "email"
  ],
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  },
  "voice": {
    "provider": "deepgram",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  },
  "firstMessage": "Hello! How can I help you today?",
  "temperature": 0.7,
  "maxTokens": 4096
}'

import requests

url = "https://api.thinnest.ai/v2/agents"
headers = {
    "Authorization": "Bearer " + "YOUR_THINNESTAI_API_KEY",
    "Content-Type": "application/json",
}

payload = {
  "name": "Sales Agent",
  "model": "gpt-4o",
  "instructions": "You are a helpful sales assistant.",
  "description": "Handles product inquiries and purchase decisions",
  "voiceEnabled": true,
  "tools": [
    "duckduckgo",
    "email"
  ],
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  },
  "voice": {
    "provider": "deepgram",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  },
  "firstMessage": "Hello! How can I help you today?",
  "temperature": 0.7,
  "maxTokens": 4096
}

response = requests.post(url, headers=headers, json=payload)
print(response.json())

const response = await fetch("https://api.thinnest.ai/v2/agents", {
  method: "POST",
  headers: {
    "Authorization": "Bearer " + "YOUR_THINNESTAI_API_KEY",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
  "name": "Sales Agent",
  "model": "gpt-4o",
  "instructions": "You are a helpful sales assistant.",
  "description": "Handles product inquiries and purchase decisions",
  "voiceEnabled": true,
  "tools": [
    "duckduckgo",
    "email"
  ],
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  },
  "voice": {
    "provider": "deepgram",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  },
  "firstMessage": "Hello! How can I help you today?",
  "temperature": 0.7,
  "maxTokens": 4096
}),
});

const data = await response.json();
console.log(data);

package main

import (
    "bytes"
    "fmt"
    "io"
    "net/http"
)

func main() {
    payload := []byte(`{
  "name": "Sales Agent",
  "model": "gpt-4o",
  "instructions": "You are a helpful sales assistant.",
  "description": "Handles product inquiries and purchase decisions",
  "voiceEnabled": true,
  "tools": [
    "duckduckgo",
    "email"
  ],
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  },
  "voice": {
    "provider": "deepgram",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  },
  "firstMessage": "Hello! How can I help you today?",
  "temperature": 0.7,
  "maxTokens": 4096
}`)
    req, _ := http.NewRequest("POST", "https://api.thinnest.ai/v2/agents", bytes.NewBuffer(payload))
    req.Header.Set("Authorization", "Bearer YOUR_THINNESTAI_API_KEY")
    req.Header.Set("Content-Type", "application/json")

    resp, err := http.DefaultClient.Do(req)
    if err != nil { panic(err) }
    defer resp.Body.Close()
    body, _ := io.ReadAll(resp.Body)
    fmt.Println(string(body))
}

Request Body

Core

name

string

required

Agent display name

model

string

default:"gpt-4o"

LLM model identifier. Use provider/model format or just model name (see LLM Models)

instructions

string

default:"You are a helpful assistant."

System prompt that defines the agent’s behavior, personality, and constraints

description

string

default:"null"

Internal description (not shown to end users)

agentType

string

default:"simple"

Agent type: simple, graph, workflow

voiceEnabled

boolean

default:"false"

Enable voice/phone call capabilities

firstMessage

string

default:"null"

Greeting message the agent speaks when a voice call starts

temperature

number

default:"null"

LLM temperature (0.0–2.0). Lower = more deterministic, higher = more creative

maxTokens

integer

default:"null"

Maximum tokens in the LLM response

endCallEnabled

boolean

default:"true"

Allow the agent to end voice calls

maxDurationSeconds

integer

default:"0"

Maximum voice call duration in seconds. 0 = unlimited

noiseCancellation

string

default:"none"

Noise cancellation: none, bvc, krisp

LLM Models

The model field accepts a string in provider/model format, or just the model name for auto-detection.

Subscription Tiers: Trial (default) — limited to gpt-4o-mini and Sarvam models. PAYG (after first top-up) — all models, OCR, and multimodal unlocked. Enterprise — all features including branding removal, teams, observability, and priority support.

Provider auto-detection rules:

Model prefix	Detected provider
`gpt-`, `o1-`, `o3-*`	`openai`
`claude-*`	`anthropic`
`gemini*`	`google`
`sarvam*`	`sarvam`
`llama-`, `mixtral-`	`groq`

Or use explicit format: "openai/gpt-4o", "anthropic/claude-3-5-sonnet-20241022".

OpenAI

Model ID	Min. Tier	Notes
`gpt-4o-mini`	Trial	Fast, cost-effective for simple tasks
`gpt-4o`	PAYG	Most capable, recommended for production
`gpt-4-turbo`	PAYG	High capability, large context
`gpt-3.5-turbo`	PAYG	Legacy, fastest
`o3`	PAYG	Reasoning model
`o3-mini`	PAYG	Lightweight reasoning
`o1`	PAYG	Advanced reasoning
`o1-mini`	PAYG	Lightweight reasoning

{ "model": "gpt-4o" }
// or explicitly:
{ "model": "openai/gpt-4o" }

Anthropic

Model ID	Min. Tier	Notes
`claude-3-5-haiku-20241022`	PAYG	Fast, affordable
`claude-3-5-sonnet-20241022`	PAYG	Best for complex tasks
`claude-3-opus-20240229`	PAYG	Highest capability

{ "model": "claude-3-5-sonnet-20241022" }
// or explicitly:
{ "model": "anthropic/claude-3-5-sonnet-20241022" }

Google

Model ID	Min. Tier	Notes
`gemini-2.0-flash-exp`	PAYG	Fast, experimental
`gemini-1.5-flash`	PAYG	Lightweight
`gemini-1.5-pro`	PAYG	High capability

{ "model": "gemini-2.0-flash-exp" }
// or explicitly:
{ "model": "google/gemini-2.0-flash-exp" }

Groq (Ultra-Fast Inference)

Groq delivers the fastest LLM inference — 200-400ms TTFT for voice agents. Recommended for low-latency voice calls.

Model ID	Min. Tier	Notes
`llama-3.3-70b-versatile`	Trial	Best for voice agents, tool calling supported
`llama-3.1-8b-instant`	Trial	Fastest, lightweight
`llama-3.1-70b-versatile`	PAYG	70B with tool calling
`qwen/qwen3-32b`	PAYG	Qwen 3, strong multilingual
`meta-llama/llama-4-scout-17b-16e-instruct`	PAYG	Llama 4 Scout
`deepseek-r1-distill-llama-70b`	PAYG	DeepSeek reasoning
`gemma2-9b-it`	PAYG	Google Gemma 2
`mixtral-8x7b-32768`	PAYG	32K context MoE

{ "model": "groq/llama-3.3-70b-versatile" }

Sarvam AI (Indian LLM)

All Sarvam models are available on Trial tier (no top-up needed).

Model ID	Min. Tier	Notes
`sarvam-m`	Trial	General purpose
`sarvam-30b`	Trial	30B parameters
`sarvam-30b-16k`	Trial	30B with 16K context
`sarvam-105b`	Trial	105B parameters
`sarvam-105b-32k`	Trial	105B with 32K context

{ "model": "sarvam-m" }
// or explicitly:
{ "model": "sarvam/sarvam-m" }

Transcriber (STT)

Speech-to-text configuration. Only used when voiceEnabled is true.

{
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  }
}

Field	Type	Required	Default	Description
`provider`	string	No	`deepgram`	STT provider
`model`	string	No	Provider default	STT model ID
`language`	string	No	`en`	Language code (`en`, `hi`, `es`, `fr`, etc.)

Deepgram (default)

Default provider. Best accuracy and lowest latency for voice agents.

Model	Description
`nova-2-conversationalai`	Best for voice agents and phone calls (default)
`nova-2-phonecall`	Optimized for phone audio
`nova-2-medical`	Medical terminology
`nova-2`	Base Nova 2 model
`nova-3`	Latest generation, highest accuracy
`nova-3-multilingual`	Multi-language support
`nova-3-medical`	Medical + Nova 3
`flux-general-en`	Recommended for voice agents — Deepgram v2 streaming API, ~200ms latency, native turn detection. Picking this auto-sets `turn_detection_mode: "stt_endpointing"`.
`flux-general-multi`	Same as flux-general-en, multilingual variant

{ "transcriber": { "provider": "deepgram", "model": "nova-2-conversationalai" } }

OpenAI

Model	Description
`whisper-1`	Whisper model
`gpt-4o-transcribe`	GPT-4o powered transcription
`gpt-4o-mini-transcribe`	GPT-4o Mini transcription

{ "transcriber": { "provider": "openai", "model": "gpt-4o-mini-transcribe" } }

ElevenLabs

Model	Description
`scribe-v2-realtime`	Real-time transcription

{ "transcriber": { "provider": "elevenlabs", "model": "scribe-v2-realtime" } }

Cartesia

Model	Description
`ink-whisper`	Fast transcription

{ "transcriber": { "provider": "cartesia", "model": "ink-whisper" } }

Sarvam (Indian Languages)

Supports Hindi, Tamil, Telugu, Kannada, Malayalam, Bengali, Marathi, Gujarati, and more.

Model	Description
`saarika:v2.5`	Latest Sarvam STT, 10+ Indian languages
`saarika:v2`	Sarvam STT v2

{ "transcriber": { "provider": "sarvam", "model": "saarika:v2.5" } }

AssemblyAI

Model	Description	Pricing
`u3-rt-pro`	Universal-3 Pro Streaming — most accurate for voice agents	$0.45/hr
`universal-streaming`	Universal Streaming — fastest English transcription	$0.15/hr
`universal-streaming-multilingual`	Universal Streaming Multilingual	$0.15/hr
`whisper-streaming`	Whisper Streaming — open-source Whisper on AssemblyAI infra	$0.30/hr
`universal-3-pro`	Universal-3 Pro (file/batch only)	$0.21/hr
`universal-2`	Universal-2 (file/batch only)	$0.15/hr

{ "transcriber": { "provider": "assemblyai", "model": "u3-rt-pro" } }

Free on Trial tier — AssemblyAI STT is included free during trial (no STT charges).

Voice (TTS)

Text-to-speech configuration. Only used when voiceEnabled is true.

{
  "voice": {
    "provider": "deepgram",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  }
}

Field	Type	Required	Default	Description
`provider`	string	No	`deepgram`	TTS provider
`voiceId`	string	No	`asteria`	Voice identifier
`model`	string	No	Provider default	TTS model (if provider has multiple)
`speed`	float	No	`1.0`	Speech speed multiplier

Deepgram Aura (default)

Default provider. Low-latency, high-quality voices.Models: aura-2 (default), auraFeatured Voices (Aura-2):

Voice ID	Gender	Accent
`aura-2-thalia-en`	Female	US English
`aura-2-andromeda-en`	Female	US English
`aura-2-helena-en`	Female	US English
`aura-2-athena-en`	Female	US English
`aura-2-aurora-en`	Female	US English
`aura-2-apollo-en`	Male	US English
`aura-2-arcas-en`	Male	US English
`aura-2-atlas-en`	Male	US English
`aura-2-aries-en`	Male	US English

Multilingual: Append language code — aura-2-thalia-es (Spanish), aura-2-thalia-de (German), aura-2-thalia-fr (French), aura-2-thalia-nl (Dutch), aura-2-thalia-it (Italian), aura-2-thalia-ja (Japanese).

{ "voice": { "provider": "deepgram", "model": "aura-2", "voiceId": "aura-2-thalia-en" } }

Cartesia

High-quality, multi-language voices with low latency.Models: sonic-3, sonic-2, sonic-turbo, sonic50+ voices available — fetched dynamically. Use the voice ID from Cartesia’s library.

{ "voice": { "provider": "cartesia", "model": "sonic-3", "voiceId": "your-cartesia-voice-id" } }

ElevenLabs

Premium voice cloning and 100+ voices.Models: eleven_flash_v2_5, eleven_flash_v2, eleven_turbo_v2_5, eleven_turbo_v2, eleven_multilingual_v2Use the voice ID from your ElevenLabs account (premade or cloned voices).

{ "voice": { "provider": "elevenlabs", "model": "eleven_flash_v2_5", "voiceId": "your-voice-id" } }

OpenAI

Models: tts-1, tts-1-hd

Voice ID	Gender
`alloy`	Neutral
`echo`	Male
`fable`	Male
`onyx`	Male
`nova`	Female
`shimmer`	Female

{ "voice": { "provider": "openai", "model": "tts-1", "voiceId": "nova" } }

Sarvam (Hindi / Indian)

Model: bulbul:v2

Voice ID	Gender
`anushka`	Female
`manisha`	Female
`vidya`	Female
`arya`	Male
`abhilash`	Male
`karun`	Male
`hitesh`	Male

{ "voice": { "provider": "sarvam", "model": "bulbul:v2", "voiceId": "anushka" } }

Rime

Models: arcana-v3, mist-v2

{ "voice": { "provider": "rime", "model": "arcana-v3", "voiceId": "your-rime-voice-id" } }

Inworld

Models: inworld-tts-1.5-max, inworld-tts-1.5-mini, inworld-tts-1-max, inworld-tts-1

{ "voice": { "provider": "inworld", "model": "inworld-tts-1.5-max", "voiceId": "your-voice-id" } }

Tools

Array of tool identifiers to attach to the agent.

{
  "tools": ["duckduckgo", "email", "calculator"]
}

Field	Type	Required	Default	Description
`tools`	string[]	No	`[]`	List of tool type identifiers

Tool ID	Description
`duckduckgo`	DuckDuckGo web search
`tavily`	Tavily AI search
`exa`	Exa neural search
`serpapi`	SerpAPI Google results
`serper`	Serper.dev search
`wikipedia`	Wikipedia lookup
`arxiv`	arXiv paper search
`hackernews`	Hacker News search
`pubmed`	PubMed medical literature

Web Scraping

Tool ID	Description
`website`	Basic URL scraping
`firecrawl`	Firecrawl web scraper
`crawl4ai`	Crawl4AI scraper
`jinareader`	Jina Reader API
`spider`	Spider web crawler
`scrapling`	Scrapling scraper
`newspaper`	Article extraction
`trafilatura`	Content extraction

Communication

Tool ID	Description
`email`	Send emails
`gmail`	Gmail integration
`slack`	Slack messaging
`discord`	Discord messaging
`telegram`	Telegram bot
`twilio`	Twilio SMS/voice
`whatsapp`	WhatsApp messaging
`sms`	SMS sending
`x`	X (Twitter)
`zoom`	Zoom meetings
`reddit`	Reddit
`sendgrid`	SendGrid email
`resend`	Resend email

Databases

Tool ID	Description
`postgres`	PostgreSQL queries
`mysql`	MySQL queries
`mongodb`	MongoDB operations
`duckdb`	DuckDB analytics
`bigquery`	Google BigQuery
`sql`	Generic SQL
`neo4j`	Neo4j graph DB
`supabase`	Supabase
`firebase`	Firebase
`airtable`	Airtable
`redshift`	AWS Redshift

Productivity

Tool ID	Description
`github`	GitHub repos/issues/PRs
`gitlab`	GitLab integration
`jira`	Jira issues
`notion`	Notion pages
`linear`	Linear issues
`trello`	Trello boards
`confluence`	Confluence docs
`asana`	Asana tasks
`monday`	Monday.com
`todoist`	Todoist tasks
`clickup`	ClickUp tasks
`calcom`	Cal.com scheduling

Google

Tool ID	Description
`googlecalendar`	Google Calendar events
`googlesheets`	Google Sheets read/write
`googlemaps`	Google Maps lookup
`googledrive`	Google Drive files

CRM & Sales

Tool ID	Description
`hubspot`	HubSpot CRM
`salesforce`	Salesforce CRM
`activecampaign`	ActiveCampaign
`apollo`	Apollo.io prospecting
`linkedin`	LinkedIn
`intercom`	Intercom support
`mailchimp`	Mailchimp campaigns
`shopify`	Shopify e-commerce
`zendesk`	Zendesk support

AI & Media

Tool ID	Description
`dalle`	DALL·E image generation
`elevenlabs`	ElevenLabs TTS
`replicate`	Replicate AI models
`fal`	Fal.ai models
`lumalabs`	Luma Labs video
`giphy`	Giphy GIFs
`unsplash`	Unsplash photos
`youtube`	YouTube data
`spotify`	Spotify data

Cloud & DevOps

Tool ID	Description
`awslambda`	AWS Lambda functions
`awsses`	AWS SES email
`s3`	AWS S3 storage
`dropbox`	Dropbox files
`onedrive`	OneDrive files
`sentry`	Sentry error tracking
`datadog`	Datadog monitoring
`msteams`	Microsoft Teams

Finance

Tool ID	Description
`yfinance`	Yahoo Finance data
`openbb`	OpenBB financial data
`stripe`	Stripe payments

Utility

Tool ID	Description
`calculator`	Math calculations
`python`	Execute Python code
`shell`	Execute shell commands
`file`	File operations
`custom_api`	Custom HTTP API calls
`openweather`	Weather data
`mem0`	Memory storage
`mcp`	Model Context Protocol
`get_user`	Get caller/user info

Knowledge Base

Attach knowledge sources by passing their IDs. Knowledge sources must be created first via the Knowledge API.

{
  "sourceIds": [1, 2, 5]
}

Field	Type	Required	Default	Description
`sourceIds`	integer[]	No	`[]`	IDs of knowledge sources to attach

Knowledge Source Types

Sources are created via the Knowledge API and can be of these types:

Type	Description
`url`	Web page — crawled and indexed
`text`	Raw text content
`file`	Document upload (PDF, DOCX, TXT)
`csv`	CSV data
`json`	JSON data
`markdown`	Markdown documents
`youtube`	YouTube video transcript
`excel`	Excel spreadsheets
`github_repo`	GitHub repository
`azure_blob`	Azure Blob Storage
`sharepoint`	SharePoint documents
`ocr`	OCR-processed documents

Voice Configuration

These sections configure voice call behavior. Only used when voiceEnabled is true.

Interruption Config

Controls how the agent handles user interruptions during voice calls.

{
  "interruptionConfig": {
    "enabled": true,
    "threshold": 0.5,
    "minSilenceDuration": 0.3,
    "minSpeechDuration": 0.1
  }
}

Field	Type	Default	Description
`enabled`	boolean	`true`	Allow users to interrupt the agent mid-speech
`threshold`	float	`0.5`	Sensitivity threshold (0.0–1.0). Lower = more sensitive
`minSilenceDuration`	float	`0.3`	Seconds of silence before detecting end of speech
`minSpeechDuration`	float	`0.1`	Minimum speech duration to trigger interruption

Recording Config

Configure call recording for voice agents.

{
  "recordingConfig": {
    "enabled": true,
    "downloadEnabled": true
  }
}

Field	Type	Default	Description
`enabled`	boolean	`false`	Enable call recording
`downloadEnabled`	boolean	`true`	Allow recording downloads via API

DTMF Config

Dual-tone multi-frequency (keypad) detection for IVR systems.

{
  "dtmfConfig": {
    "enabled": true,
    "ivrDetection": true,
    "menuEnabled": true
  }
}

Field	Type	Default	Description
`enabled`	boolean	`false`	Enable DTMF keypad detection
`ivrDetection`	boolean	`false`	Detect IVR system prompts
`menuEnabled`	boolean	`false`	Enable DTMF menu navigation

Silence Config

Control agent behavior during silence and unresponsive callers.

{
  "silenceConfig": {
    "unresponsiveTimeoutSeconds": 30,
    "unresponsiveFinalSeconds": 15,
    "fillersEnabled": true,
    "fillerPhrases": ["Let me think about that...", "One moment please..."]
  }
}

Field	Type	Default	Description
`unresponsiveTimeoutSeconds`	integer	`30`	Seconds before first silence prompt
`unresponsiveFinalSeconds`	integer	`15`	Seconds before ending the call
`fillersEnabled`	boolean	`false`	Use filler phrases during processing
`fillerPhrases`	string[]	`[]`	Custom filler phrases

Multilingual Config

Enable multi-language support in voice calls.

{
  "multilingualConfig": {
    "enabled": true,
    "primaryLanguage": "en"
  }
}

Field	Type	Default	Description
`enabled`	boolean	`false`	Enable multilingual detection
`primaryLanguage`	string	`en`	Primary language code

Call Summary Config

Automatically generate a summary after each voice call ends.

{
  "callSummaryConfig": {
    "enabled": true,
    "prompt": "Summarize the call including: key topics discussed, action items, and customer sentiment."
  }
}

Field	Type	Default	Description
`enabled`	boolean	`false`	Enable post-call summaries
`prompt`	string	`""`	Custom summarization prompt

Webhook Config

Receive real-time event notifications for voice calls.

{
  "webhookConfig": {
    "url": "https://your-server.com/webhook",
    "events": ["call_started", "call_ended", "message_received"],
    "retryCount": 3
  }
}

Field	Type	Default	Description
`url`	string	`""`	Webhook endpoint URL
`events`	string[]	`[]`	Events: `call_started`, `call_ended`, `message_received`, `error_occurred`
`retryCount`	integer	`3`	Number of retry attempts on failure

Graph Data

Workflow configuration for graph or workflow agent types. Used with the visual flow editor.

Graph Data Structure

{
  "agentType": "graph",
  "graphData": {
    "nodes": [
      {
        "id": "trigger-1",
        "type": "trigger",
        "data": { "label": "Start" }
      },
      {
        "id": "agent-1",
        "type": "agent",
        "data": {
          "label": "Support Agent",
          "model": "gpt-4o",
          "instructions": "Handle customer support queries."
        }
      }
    ],
    "edges": [
      {
        "id": "e1",
        "source": "trigger-1",
        "target": "agent-1"
      }
    ]
  }
}

Node Types:

Type	Description
`trigger`	Entry point — starts the workflow
`agent`	LLM agent node
`team`	Multi-agent team node
`condition`	Conditional branching
`tool`	Tool execution
`llm`	Direct LLM call
`http-request`	External API call
`python`	Python code execution
`variable`	Set/get variables
`memory`	Memory operations
`delay`	Wait/pause
`webhook`	Webhook trigger
`approval`	Human approval gate
`loop`	Loop iteration

Full Example — Voice Agent with Tools

{
  "name": "Sales Support Agent",
  "model": "gpt-4o",
  "instructions": "You are a sales support agent for Acme Corp. Help customers with product questions, pricing, and order status.",
  "description": "Handles inbound sales calls",
  "voiceEnabled": true,
  "firstMessage": "Hi there! Thanks for calling Acme Corp. How can I help you today?",
  "temperature": 0.7,
  "maxTokens": 4096,

  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  },

  "voice": {
    "provider": "deepgram",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  },

  "tools": ["duckduckgo", "email", "calculator", "googlesheets"],
  "sourceIds": [1, 3],

  "interruptionConfig": { "enabled": true, "threshold": 0.5 },
  "silenceConfig": {
    "unresponsiveTimeoutSeconds": 20,
    "fillersEnabled": true,
    "fillerPhrases": ["Let me check that for you...", "One moment please..."]
  },
  "recordingConfig": { "enabled": true },
  "callSummaryConfig": {
    "enabled": true,
    "prompt": "Summarize: topics discussed, products mentioned, next steps, and customer sentiment."
  },
  "webhookConfig": {
    "url": "https://api.acme.com/call-events",
    "events": ["call_started", "call_ended"]
  },

  "maxDurationSeconds": 900,
  "noiseCancellation": "krisp",
  "endCallEnabled": true
}

Full Example — Simple Chat Agent

{
  "name": "FAQ Bot",
  "model": "gpt-4o-mini",
  "instructions": "You are a helpful FAQ assistant. Answer questions based on the provided knowledge base. If you don't know the answer, say so.",
  "tools": ["duckduckgo"],
  "sourceIds": [2]
}

Response `201`

Returns the full agent object with nested configuration (see Get Agent for complete response schema).

{
  "id": "ag_c47e7c97_b2f2",
  "name": "Sales Support Agent",
  "createdAt": "2026-03-07T10:00:00Z",
  "updatedAt": null,
  "model": {
    "provider": "openai",
    "model": "gpt-4o",
    "temperature": 0.7,
    "maxTokens": 4096,
    "instructions": "You are a sales support agent for Acme Corp..."
  },
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-conversationalai",
    "language": "en"
  },
  "voice": {
    "provider": "deepgram",
    "model": "aura-2",
    "voiceId": "aura-2-thalia-en",
    "speed": 1.0
  },
  "tools": [
    { "name": "duckduckgo", "type": "duckduckgo", "enabled": true },
    { "name": "email", "type": "email", "enabled": true },
    { "name": "calculator", "type": "calculator", "enabled": true },
    { "name": "googlesheets", "type": "googlesheets", "enabled": true }
  ],
  "knowledgeBase": {
    "sources": [
      { "id": "1", "name": "Product Catalog", "type": "file", "status": "ready" },
      { "id": "3", "name": "Pricing FAQ", "type": "text", "status": "ready" }
    ]
  },
  "voiceEnabled": true,
  "agentType": "simple",
  "firstMessage": "Hi there! Thanks for calling Acme Corp. How can I help you today?",
  "endCallEnabled": true,
  "maxDurationSeconds": 900,
  "noiseCancellation": "krisp"
}

Voice configuration

For voice agents, pass a voice object with the cascaded TTS picker, Speech-to-Speech settings (Gemini Live or OpenAI Realtime), noise cancellation, and audio ambience. See the dedicated Voice Config Reference for every field, default, and accepted values.

Errors

Code	Description
`401`	Missing or invalid API key
`403`	Agent limit reached for your plan
`422`	Invalid request body

Get Agent Update Agent

Introduction

Agents

Chat

Calls

Sessions

Phone Numbers

Tools

Knowledge

Campaigns

Webhooks

BYOK Keys

Request Body

Core

LLM Models

Transcriber (STT)

Voice (TTS)

Tools

Knowledge Base

Voice Configuration

Graph Data

Full Example — Voice Agent with Tools

Full Example — Simple Chat Agent

Response `201`

Voice configuration

Errors

​Request Body

​Core

​LLM Models

​Transcriber (STT)

​Voice (TTS)

​Tools

​Knowledge Base

​Voice Configuration

​Graph Data

​Full Example — Voice Agent with Tools

​Full Example — Simple Chat Agent

​Response 201

​Voice configuration

​Errors

Request Body

Core

LLM Models

Transcriber (STT)

Voice (TTS)

Tools

Knowledge Base

Voice Configuration

Graph Data

Full Example — Voice Agent with Tools

Full Example — Simple Chat Agent

Response `201`

Voice configuration

Errors