Pricing
You pay for the memory layer — persistent memory, the coherence gate, and overnight learning that compounds as your agent works. Khwan never runs a model, so there are no token allowances and no markup: you run your own model and pay your provider directly.
The Khwan layer
The product itself — persistent memory, the coherence gate, overnight learning, and isolated cores. This is what your subscription buys, and its value grows the more you use it.
Your model
You run it — any provider, your key. Khwan builds the context and learns from the answer, but never calls your model, meters its tokens, or marks it up. Not billed by us.
Free
Try the memory layer on one core.
Includes
- 1 core (isolated brain)
- Weekly synthesis · 30-day retention
- Playground & debug
- 1 seat · 2 req/s
Starter
For solo builders shipping their first agent.
Includes
- Everything in Free
- 5 cores · nightly synthesis
- 6-month retention
- Per-user sub-brains
- 10 req/s · priority support
Pro
For teams running agents in production.
Includes
- Everything in Starter
- 25 cores · nightly + on-demand synthesis
- 18-month retention · 5 team seats
- 50 req/s
- Usage analytics
Scale
On-prem, single-tenant, and compliance.
Includes
- Everything in Pro
- Unlimited cores · custom synthesis
- On-prem / self-hosted · SLA
- 200+ req/s
- Custom retention & data controls
Billed monthly, cancel anytime. Rate limits are requests/second (burst-tolerant); every plan pools its usage across cores.
Every plan is model-agnostic: you run your own model and pay your provider directly. Khwan never meters or marks up model usage.