Skip to main content
Remem is a memory API built for AI agents — it gives your product the ability to remember what users say, prefer, and do across every session. With a single remember() call your agent stores what it learns, and a single context() call retrieves the most useful memories at the start of each conversation. These docs cover everything from getting your first API key to integrating Remem into a production LangGraph agent.

Quick Start

Store your first memory and retrieve it in under 5 minutes

Authentication

Get your API key and make your first authenticated request

API Reference

Full endpoint docs for every Remem memory operation

LangGraph Guide

Add persistent memory to any LangGraph agent in minutes

​
Why Remem

Most vector stores return the most similar memory — Remem returns the most useful one. Hybrid scoring weights every result by 70% semantic relevance, 20% recency, and 10% importance, so stale or trivial memories never crowd out what actually matters. Built-in duplicate detection keeps your memory store clean automatically, and the score_detail field on every result makes the ranking fully transparent — no black box.

​
How It Works

1

Get your API key

Sign up at dev.remem.online — your key arrives in your inbox within 60 seconds. It starts with rm_live_.
2

Install the SDK

3

Store a memory

4

Retrieve at session start

​
Explore the Docs

Memory Types

Episodic, semantic, and summary — when to use each

Hybrid Scoring

How Remem ranks memories by relevance, recency, and importance

Error Handling

Handle errors gracefully so your agent never crashes in production