
Real-time AI infrastructure for characters that See, Speak, & Remember
import { EstuaryClient } from "@estuary-ai/sdk";
const client = new EstuaryClient({
serverUrl: "https://api.estuary-ai.com",
apiKey: "est_your_api_key",
characterId: "your-character-id",
playerId: "player-1",
});
client.on("botResponse", (r) => console.log(r.text));
await client.connect();
client.sendText("Hello!");Memory that compounds. Estuary builds a knowledge graph from every interaction. Characters recall what matters across sessions, platforms, and users.
Sam × Luna
Character memory graph
Deploy multiple characters that share context, riff off each other, and coordinate autonomously.
Give every character a distinct voice, personality, and provider stack. Ship real-time voice without building the pipeline.
Per-character voices
Clone from a sample, pick from a library, or bring your own. Every character deserves a voice that fits.
Natural interruption
Turn-taking logic that handles interruption out of the box.
Swappable provider stack
Mix STT, LLM, and TTS providers per character. Optimize for cost, latency, or quality.
Define personality, craft a voice, pick your model stack. One dashboard for everything that makes a character feel real.



Characters follow your users across devices and remember every conversation.

Validated by academia and the builder community.

MIT Reality Hack 2026 - Best AR Utility (Qualcomm) + 2nd Place XR Health & Wellness Research & Design Track
Built in 48 hours. A smartglass AI character that remembers a user's life context: names, tasks, routes home and retrieves it through natural conversation. Powered by Estuary's persistent memory and real-time voice pipeline running on RayNeo X3 Pro glasses.
Learn More
MIT Reality Hack 2025 - Best Use of Apple Vision Pro
Built in 60 hours. Multiple users share a space with AI pets that learn about their owners through conversation, then autonomously connect compatible owners by initiating play sessions with each other. Powered by Estuary's multiuser support and multiagent coordination on Apple Vision Pro.
Learn More
ACM Human Factors in Computing Systems (CHI) Case Study 2025
Interviews with leading AI agent researchers validated the problem Estuary addresses: teams building AI characters face a choice between patching together generic APIs or building custom infrastructure from scratch. Researchers voiced that platform & deployment agnosticity, persistent memory, and multimodal perception - Estuary's core capabilities - are the most critical unmet needs.
Read Paper
ACM Intelligent Virtual Agents (IVA) 2024
Introduces Estuary's core architecture: a modular, off-cloud framework for building multimodal AI characters with low-latency real-time interaction. Demonstrates how Estuary's design enables characters to run entirely off-cloud for full data privacy while maintaining real-time response across text, audio, and vision modalities.
Read Paper