Databaset

AI memory infrastructure

Persistent memory for AI apps
without the infra sprint

Databaset remembers what your users say, finds it by meaning when it matters, and hands your model ready-to-use context. No vector DB sprint. No chunking scripts. No rebuilding the same pipeline on every project.

Start for FreeLive Demo

Free tier includes 10,000 memories/month. No credit card required.

Stored memories3 indexed
Building SaaS in Next.js
Prefers PostgreSQL
Dark mode always on

Recall query

"tech stack"

Context for LLM

Waiting for recall…

3

Lines to integrate

One npm install, one API key

< 50ms

Recall latency

p95 across all regions

99.9%

Uptime SLA

On Growth and Enterprise plans

∞

Users per API key

Isolated memory per userId

Works with every AI model and framework

OpenAIAnthropicVercelSupabaseLangChainNext.js

Live demo

See AI memory in action, live

Store a few facts, ask a question, and watch Databaset recall relevant context before answering. No signup required.

StoreRecallAnswer
Memories0
0/2000

Stored

  • Loading...
Add a few facts on the left, then ask me anything. I'll recall what matters before I answer.
Platform

One API. Full memory stack.

Stop stitching Pinecone, chunking scripts, and retrieval glue. Ship memory like you ship auth.

01

Storage

Auto chunking and embeddings

Send raw text. We split on sentence boundaries, embed with production models, and store vectors in pgvector. You never touch chunk size or embedding config.

< 30msmedian store latency

02

Retrieval

Semantic recall with reranking

Queries match by meaning, not keywords. Recency, memory type, and contradiction handling are built in. No extra config required.

03

Dashboard

See every memory per user

Search, filter, inspect similarity scores, and delete. Debug what your model actually knows about each user.

100%of memories visible in UI

Security

Encrypted and isolated

Per-user isolation, hashed API keys, TLS 1.3, AES-256 at rest. SOC2 path on Enterprise.

Scale

One key, unlimited users

Multi-tenant by design. Separate prod and staging with appId namespaces.

Quickstart

Four steps to production

Install, store, recall, ship. No vector DB project required.

1

Install

One package. No other dependencies.

bash
npm install @databaset/sdk
2

Store

Send any text. We handle chunking, embedding, and storage automatically.

javascript
await memory.store(userId, "User prefers dark mode")
3

Recall

Semantic search returns the most relevant memories.

javascript
const context = await memory.recall(userId, userMessage)
4

Use with any AI

Inject context into Claude, GPT-4, Gemini, or any model.

javascript
const reply = await anthropic.messages.create({  system: `You know this about the user: ${context}`,  messages: [{ role: 'user', content: userMessage }]})

Pricing

Simple, predictable pricing

Start free, grow when you're ready. No surprise bills.

Free

Perfect for side projects and prototypes

$0/month
  • 10,000 memories stored
  • 1,000 recalls/month
  • 1 app
  • 7 day retention
  • Community support
  • Full API access
Start free, no credit card
Most Popular

Starter

For production apps with real users

$29/month
  • 500,000 memories stored
  • 50,000 recalls/month
  • 10 apps
  • Unlimited retention
  • Email support
  • Dashboard analytics
  • Webhooks
Start Starter

Growth

For teams scaling AI products

$99/month
  • 5,000,000 memories stored
  • 500,000 recalls/month
  • Unlimited apps
  • Unlimited retention
  • Priority support
  • Advanced analytics
  • Custom metadata
  • SLA guarantee
Start Growth

Enterprise

Custom limits and dedicated support

Custom
  • Unlimited everything
  • Self-hosted deployment option
  • Dedicated support engineer
  • Custom contract and billing
  • SOC2 compliance
  • HIPAA available
Contact Sales

FAQ

Get started free

Your users deserve memory that lasts

10,000 memories free every month. SDK, dashboard, and docs included. No credit card.

Start for FreeRead docs