Introduction

Saaya is a real-time multimodal AI agent platform. One config, every channel — voice, video, chat, and WhatsApp — wired up with the LLMs, voices, and avatars you actually want to ship.

What is Saaya?

Saaya is the runtime that takes your AI agent from a single prompt to a production conversation. You define behaviour. We handle the parts that silently break in production — turn-taking, citations, retries, fallbacks, observability, deployment.

Every agent ships with:

• Multimodal output: voice, video, chat, and WhatsApp from one config.
• Provider-agnostic LLM, STT, TTS, and avatar layers — switch in seconds, no lock-in.
• RAG over your knowledge bases with strict citations or controlled escalation.
• Versioning, rollback, and per-call config caching for sub-second cold starts.
• Full observability — transcripts, sentiment, latency breakdowns, tool-call inspection.

How Saaya fits

Saaya sits between your product and the conversation layer. Your code talks to the Saaya REST API or SDK; Saaya runs the agent against your callers in real time.

Your first call

Most teams ship their first agent in under fifteen minutes. The shortest path is the single-prompt agent — one config, every channel — with the @saaya/sdk in your favourite runtime.

import { Saaya } from "@saaya/sdk";

const saaya = new Saaya({ apiKey: process.env.SAAYA_API_KEY });

const agent = await saaya.agents.create({
  name: "Pipeline Concierge",
  prompt: "You're Maya, a friendly inbound qualifier for Saaya.",
  voice: { provider: "elevenlabs", voiceId: "rachel" },
  llm:   { provider: "anthropic", model: "claude-opus-4-7" },
});

console.log(`Agent ready: ${agent.id}`);

Where to next

From here you can keep going down whichever path fits the way you build:

• Quickstart — the five-minute path from zero to a live agent.
• API reference — every endpoint, schema, and code sample.
• MCP endpoint — let Claude, ChatGPT, and Cursor query Saaya directly.
• Changelog — what shipped recently.