Add memory to your AI product

Give your users persistent memory — without building five storage systems.

error_outline

Without memory

Your AI product needs per-user memory, and building it properly means vector search plus a knowledge graph plus episodic storage plus consolidation workers plus tenant isolation — a distributed-systems project bolted onto your actual roadmap.

check_circle

With Rekall

Rekall is the memory backend: SDKs in TypeScript, Python, and Go, tenant isolation enforced per-store, granular scopes for least-privilege keys, and the whole 5-layer architecture behind one API. Your users get memory; you ship your product.

How it works

Integrate the SDK

A few lines per interaction: store what happened, recall what matters. The layering (vector, graph, episodic, procedural) is handled for you.

Isolate your tenants

Org scoping and membership checks run at every store — one user’s memory can never surface in another’s recall.

Scale on the platform

Consolidation, decay, and graph maintenance run as managed background work; quotas and rate limits keep costs predictable.

See it in action

Per-user memory in an afternoon

import Rekall from '@rekall/agent-sdk'
const rekall = new Rekall({ apiKey: 'rk_...' })

// Inside your product's chat handler
await rekall.memories.create({
  content: userMessage,
  hiveId: `user-${user.id}`,   // hard isolation
})
const context = await rekall.memories.search({
  query: userMessage,
  hiveId: `user-${user.id}`,
})