Introduction

RetainDB gives your AI application persistent memory. Users tell it things once — it remembers across every session, automatically.

The full integration is three lines:

import { RetainDB } from "@retaindb/sdk";

const db = new RetainDB({ apiKey: process.env.RETAINDB_KEY });

const { response } = await db.user(userId).runTurn({
  messages,
  generate: (ctx) => llm.chat(ctx),
});

Every runTurn call retrieves the user's relevant memories, injects them into your LLM, generates a response, and stores the conversation for future use — without any extra code.

What RetainDB does

Where to start

• Fastest — Setup Wizard: guided, working in minutes.
• Manual proof — 5 minute quickstart: make one API call by hand.
• Start building — SDK quickstart: the full integration pattern.
• MCP client — MCP setup: for Claude Desktop, Claude Code, or Cursor.

The mental model

RetainDB has two main surfaces:

1. Memory — per-user facts, preferences, events, and instructions that persist indefinitely.
2. Index — project-scoped knowledge from documents, repos, and websites you ingest.

A typical turn:

1. User sends a message.
2. runTurn retrieves relevant memories for that user.
3. Memories are injected into the LLM system prompt automatically.
4. The response is returned, and the conversation is stored for next time.

That loop — retrieve, inject, generate, store — is what makes your app smarter with every interaction.

Next step

Go to SDK quickstart to see the complete pattern, or 5 minute quickstart to verify the API shape first.