Open core. Per call.

Pay for what you ask, not who you are.

The client is open source. The benchmark is open source. The hosted multi-tenant brain is per-API-call. No seat licenses, no tier games.

OPEN SOURCE

Self-host

The OSS client, schemas, and benchmark suite. Run it yourself on your own four stores.

Free
MIT-licensed client
  • Full ingest pipeline
  • Multi-signal recall
  • Benchmark suite
  • Community support
View on GitHub
ENTERPRISE

Dedicated cortex

Dedicated infrastructure, custom regions, compliance reviews, audit-grade isolation.

Custom
Annual commit
  • Dedicated cluster
  • Custom regions
  • SOC 2 collateral
  • Named support
Talk to us
The cost story

Constant tokens. Compounding memory.

Most memory layers grow your context window until the model breaks. LeapMemory returns ranked, scored memory. Cost stays flat as memory compounds.

month 1
month 6
month 12
LM y1
LM y3

Stuff-the-context vs ranked recall.

Stuffing history into the context window scales linearly with conversation length. LeapMemory returns a ranked, bounded slice. Your prompt size stays the same in week one and year three.

Ready to try it?

Join the waitlist for hosted access. Or get the open-source client and self-host.