agent-memory-sdk

TypeScript SDK for building AI agents with automatic, scoped, persistent memory.

Full project documentation: github.com/gharibyan/agent-memory.

Install

npm install agent-memory-sdk

pnpm add agent-memory-sdk
yarn add agent-memory-sdk

Quick Start

import { createAgent, openai } from "agent-memory-sdk"

const agent = createAgent({
  model: openai("gpt-5")
})

const result = await agent.generate({
  userId: "user_123",
  threadId: "thread_123",
  operationId: "op_weekly_update",
  messages: [
    { role: "user", content: "Remember that I prefer concise weekly reports." }
  ]
})

console.log(result.text)

First-party helpers include openai(), anthropic(), gemini(), and xai(). Use openAICompatible() for custom chat-completions endpoints.

If userId is omitted, memory is stored in the shared default scope. Local memory defaults to .memory/memory.json; use sqliteMemory() for .memory/memory.sqlite or postgresMemory() for Postgres with automatic pgvector migrations.

Data Migration

Seed existing app data by passing normalized memories or events to the configured memory store:

await agent.memory.migrate({
  userId: "user_123",
  data: [
    {
      type: "preference",
      content: "User prefers concise weekly reports.",
      confidence: 0.9,
      importance: 0.8
    }
  ]
})

The SDK only validates and stores mapped input. Your app owns where the data comes from and how it is mapped. Use events and sourceEventIds when you want imported memories linked to imported history, and mode: "skipExisting" to leave matching memories unchanged.