Memory infrastructure for AI products

Give your AI
a memory that lasts.

One API key. Each person keeps their own context across sessions. Scroll for a short code example.

TypeScript · any AI model · one memory per user

Where Memonest fits

Your users
User activity and actions in your app
Memonest
Saves and loads context for each user ID
Your AI
Sees memory and profile in the prompt
Open console

How it works

Save once, reuse before every reply

Each user has an ID. You send user activity or notes; we store them and find the best matches plus profile lines for your next prompt—no vector database for you to run.

Save

From activity or your app

Find

By meaning, per user

Use

Profile + memories in the prompt

See the simple flow →

agent.ts
import MemoryClient from "@memonest/sdk";

const client = new MemoryClient({
  apiKey: process.env.MEMONEST_API_KEY
});

// Pull useful facts from recent activity
await client.auto({
  userId: "user_492",
  messages: [ { role: "user", content: "I prefer dark mode and Python." } ],
  maxMemories: 5,
});

// Memories + profile text for your prompt
const { memories, static: profile } = await client.findPersonalized({
  userId: "user_492",
  query: "theme preferences",
  limit: 5,
});

Go live fast: get a key, add our SDK, three calls.

Simple flow

Your app, Memonest, then your AI

After each user action: call auto (learn from activity history) or add (save a fact). Before the AI answers: call findPersonalized, then add the returned static and memories to your prompt.

1

Save

Send activity data or facts—we store and index them for you.

2

Find

Ask in plain language; we return the best matches for one user.

3

Reply

Put those results in your prompt so the AI answers with context.

Why Memonest

No vector database to run

We host search and storage—you focus on your product.

Three clear API calls

Learn from activity (auto), fetch context (findPersonalized), or save a fact (add).

Quick to add

Grab an API key and plug it into your existing agent loop.

Free to start

Use the free tier; upgrade when you need more usage.

SDK

Three API calls

No tuning search indexes. Learn from activity, pull context with profile, or save a fact by hand.

Call 01
auto

Reads recent messages and saves useful memories for you.

POST /v1/memories/auto
{ "user_id": "…", "messages": [...] }
+
Core step
Call 02
findPersonalized

Gets the best-matching memories plus profile lines for your prompt.

POST /v1/memories/search-personalized
{ "static": [...], "memories": [...] }
+
Call 03
add

Saves something you already know about the user.

POST /v1/memories
{ "user_id": "…", "content": "…" }

Stop starting from zero context

Try free first. Same simple SDK when you grow.

Build yourself vs Memonest
Topic
Build yourself
With Memonest
What you build
Login, app, plus a whole memory system
Login, app, plus Memonest
Memory per user
You design and code it
Built in—tied to each user ID
Servers and search
Vector DB, servers, background jobs
We run the API; you use the SDK
Time to first version
Often two or three weeks
About ten minutes
Ongoing work
Scaling, updates, cost control
We handle the heavy lifting

Simple pricing

Start free. Pay more only when you need higher limits.

TIER // 01

Hobby

$0/mo

Try ideas and side projects at no cost.

  • Up to 1M tokens / month
  • Up to 10,000 searches / month
  • All API features included
  • Community support
  • No credit card
Open console — start free
Most popular
TIER // 02

Production

$18/mo

For live apps with more traffic and higher limits.

  • Up to 3M tokens / month
  • Up to 100,000 searches / month
  • Room for growth and heavier use
  • Priority support
Upgrade in console — $18/mo