The Developer-First Voice Agent Framework

Build AI Voice AgentsLike a Software Engineer

Production-ready voice agents in TypeScript. Skip the infrastructure—just write the logic.

TypeScript-First API

IDE autocomplete, type safety, and modern tooling

Engineer-Grade Tooling

CLI, local dev server, and production observability

Infrastructure as Code

Declare providers, fallbacks, and scaling in config

Start Building View Documentation

Built for Engineering Teams

Enterprise-ready with developer experience as priority one

# Install the framework
npm install -g @interactkit/sdk

# Initialize a production-ready project
interactkit create-app

# Check out your new app
cd my-app && code .

Code the Brain, We Handle Everything Else

InteractKit separates configuration from execution. You write the bot logic, our managed runtime handles telephony, audio streaming, LLM orchestration, and infrastructure.

Your Code (Minimal)

Just define configuration & logic

LLM Configuration
Declare which models to use (Anthropic, OpenAI, etc.)
Tools as Class Methods
Define tools with simple async methods, no JSON schemas
Single Configuration File
All settings in one place: LLM, TTS, STT, prompts, tools
Event Handlers (Optional)
Subscribe to lifecycle events when needed

Our Managed Runtime (Everything Else)

We handle the complex infrastructure

Telephony & WebRTC Management
Full audio streaming pipeline with auto-scaling
Complete Orchestration Loop
STT → LLM → Tool Calling → TTS → Audio streaming
Automatic Session Management
State, transcripts, logging, and fault tolerance
High-Availability Infrastructure
Multi-region deployment with automatic failover

The Complete Voice Agent Stack, Abstracted

InteractKit abstracts away the entire real-time voice pipeline so you can focus on building intelligent agents, not infrastructure.

Managed Telephony & Streaming

WebRTC, SIP, and PSTN connections with automatic audio streaming. No WebSocket or low-level audio handling required.

LLM Orchestration Layer

Automatic prompt construction, tool calling routing, and multi-model support. Just declare which providers to use.

Tool Execution Engine

Tools as simple async methods. Automatic validation, execution, and result injection back into conversation.

Real-time Conversation Loop

Full STT→LLM→TTS pipeline with automatic session tracking and state management.

Built-in Session Management

Automatic transcripts, session lifecycle events, and persistent storage for recordings and state.

Serverless Execution Model

Your bot runs in our managed runtime, not your serverless functions. Keep your infrastructure minimal.

Build Complex Bots with Simple Code

Define your entire bot in one TypeScript class. No audio streaming, no WebSocket handling, no complex orchestration code.

No infrastructure code

Automatic tool execution

Built-in error handling

bot.ts

@InteractKitBotProto({ name: "SupportBot" })
export class SupportBot extends InteractKitBot {
  async initialize() {
    return {
      llm: [new AnthropicLLM()],
      tts: [new ElevenLabsTTS()],
      prompt: "You're a helpful support agent...",
      toolKits: [new SupportTools()]
    };
  }

   ... inside the SupportTools class ...

  // Tools are simple async methods
  @ToolProto({ name: "getUserInfo" })
  async getUserInfo(userId: string) {
    // SDK handles execution & LLM integration
    return db.users.find(userId);
  }
}

Deploy Anywhere, Scale Everywhere

Your bot runs in InteractKit's managed runtime. Deploy a minimal Express app to any platform while we handle the heavy lifting.

Tiny Deployment Footprint

Deploy only a thin integration layer. No audio processing or LLM logic runs in your infrastructure.

Automatic Scaling

InteractKit handles concurrency, load balancing, and auto-scaling. Your bot scales to zero when idle.

Production-Ready

Built-in monitoring, logging, analytics, and enterprise-grade security out of the box.

Choose Your Stack, We Handle Integration

Mix and match providers. InteractKit handles all integrations automatically.

Anthropic

LLM

OpenAI

LLM

ElevenLabs

TTS

Deepgram

STT

Twilio

Telephony

AWS S3

Storage

Redis

Cache

PostgreSQL

Database

View all providers and integrations

Start Building in Minutes, Not Weeks

Skip the infrastructure complexity. Define your bot's brain, we'll handle the nervous system, voice, and scalability.

Build Your First Bot Schedule a Demo

Free tier includes 1 free cold agent for lifetime. • No credit card required