Products

One memory & storage stack. Open source, top to bottom.

From prototype to production — the open-source MemoryLayer family gives every AI application a durable, queryable memory layer that scales with you.

Apache-2.0 & AGPL-3.0 Same SDK across all tiers Self-host anywhere
Apache-2.0

MemoryLayer

  • Self-hostable memory API — Apache 2.0, no lock-in
  • SQLite backend: zero-config, runs completely offline
  • Python, TypeScript & Go SDKs + MCP server for Claude, Cursor, and more
  • Hybrid BM25 + vector retrieval, knowledge graph, automatic fact extraction
Learn more
AGPL-3.0

MemoryLayer Enterprise

  • PostgreSQL + pgvector: durable, scalable production backend
  • Everything in MemoryLayer, at production scale — open source and self-hosted
  • Collections, datasets, document chat, and enterprise data connectors
  • RBAC, audit trails, SSO/OIDC, and multi-tenant isolation
Learn more
AGPL-3.0

MemoryLayer Storage

  • Content-addressed dedup + compression — never store the same bytes twice
  • S3-compatible gateway: drop-in for any existing S3 client
  • POSIX filesystem (FUSE) on the same deduped substrate
  • Pluggable backends: S3, RustFS, local FS — self-host anywhere
Learn more
One SDK

Start with MemoryLayer. Move to Enterprise when you're ready.

The MemoryLayer client SDK is identical across every edition. Swap from SQLite to Postgres — your application code doesn't change. The upgrade is purely operational.

app.py
from memorylayer import SyncMemoryLayerClient

# MemoryLayer: SQLite, zero config
url = "http://localhost:61001"

# Enterprise: just a different URL
url = "https://ml.internal"

# Your code is identical either way
with SyncMemoryLayerClient(base_url=url) as ml:
    ml.remember(content)

Ready to give your AI a memory?

Everything is open source. Start with MemoryLayer and grow into Enterprise and Storage when your workload demands it.