Today we're opening MemNexus to a small group of developers as a gated preview. If you're reading this, you probably got an invite. Here's what you need to know.

What is MemNexus?

MemNexus is a persistent memory layer for AI assistants. It stores what you and your AI tools learn together — decisions, patterns, context — and makes it available across conversations, across tools, and across time.

The problem is simple: AI assistants forget everything when a conversation ends. You end up re-explaining your stack, your conventions, your project history, over and over. MemNexus fixes that.

How it works

MemNexus stores memories in a graph database with semantic search. When you save a memory, we generate vector embeddings and store the relationships between topics, entities, and facts. When you or your AI assistant search, we run a hybrid search — combining vector similarity with keyword matching — and return the most relevant results.

The fastest way to get started is with the CLI:

npm install -g @memnexus-ai/cli
mx auth login --interactive

Then start saving and searching memories:

mx memories create \
  --content "Our API uses rate limiting at 100 req/min per key, based on P95 usage data." \
  --topics "api,rate-limiting,decision"

mx memories search --query "what's our rate limit"

The CLI works with any command-line AI assistant. Tell Claude Code, GitHub Copilot, or Cursor to run mx commands — they can save and retrieve memories on your behalf without any extra setup.

What's in the preview

You get the full product:

CLI (@memnexus-ai/cli) — Create, search, and manage memories from the terminal
Dashboard — Browse memories, manage API keys, view usage
REST API — Full API access for custom integrations
TypeScript SDK (@memnexus-ai/mx-typescript-sdk) — Programmatic access
Free tier — 200 memories, 60-day retention
Pro tier — Unlimited (simulated billing via Stripe test mode)

What we're looking for

This isn't a beta — everything should work. We're validating three things:

Do people actually want this? Will you come back after the first session?
Where do you get stuck? Can you go from invite to first memory without hand-holding?
Does it hold up? Bugs, performance issues, things that feel wrong.

Your honest feedback is the most valuable thing you can give us.

How to give feedback

Real-time help — Join our Discord
Bug reports and feature requests — GitHub Issues
Security concerns — Email [email protected] directly

Get started

If you have your invite code, head to the welcome guide. You'll be up and running in under 10 minutes.

Thanks for being part of this. Your feedback shapes what we ship.