Pricing and licensing

Flat per hospital tenant. Never per agent.

Not per seat, not metered. Run one agent or twenty on the same tenant and the price does not move. A tenant is one hospital or health system your product serves.

Platform subscription

One flat fee per hospital tenant, per month

The multi-tenant platform deployed in your AWS, GCP, Azure or Databricks account. Unlimited agents, unlimited calls, every skill in the open catalog.

Never per seat, never metered
Roughly a tenth of what a clinical-AI vendor charges a hospital
Shared on the first call

Book the first call

Implementation and support

One-time plus annual support

White-glove first-tenant deployment into your own cloud, then enterprise support for the life of the platform.

SLA, upgrades, security patches
Rotation runbooks for keys and entitlements
A named contact who knows your stack

Custom agents and specialized skills

Scoped engagement fixed price per project, plus premium skills per tenant

When your specialty needs more than the catalog: we author and clinically review the skills, build the agentic workflow around them, test against synthetic cohorts, and deploy into your tenant.

Specialized skills written in SKILL.md and reviewed by clinicians
Custom agentic workflows: interactive, headless, or cohort-scale
Build, test on synthetic patients, deploy to your tenant, hand over
Premium skills and accelerators attach per tenant, never renegotiated

Why flat.

Per-agent pricing punishes the thing we want you to do, which is run more of your product on one guardrailed layer.

Your hospital contract is the big number. Kartha is a rounding error next to it, and it stays that way as you add agents, add users and add calls. The only thing that changes your bill is a new hospital going live, which is also the moment you have new revenue.

Because the platform runs in your account with your keys, there is no usage we could meter without sitting in the PHI path. We would rather not be there, and neither would your security team.

What a tenant is

One hospital or health system that your product serves, with its own EMR connection, its own entitlement and its own isolation inside your deployment. A ten-hospital health system on one Epic instance is one tenant.

How licensing works.

Open-source core

LangCare, the FHIR MCP server, is MIT-licensed and free. Kartha's commercial platform is the tenancy, guardrails, skills, workflows and support on top.

Signed entitlement, offline

A compact signed file lists your tenants and premium skills. The binary verifies it with an embedded public key. No runtime license server, no phone-home.

Never blocks a clinical call

If an entitlement is missing or expired the platform degrades to grace: premium skills flag, your team is alerted, FHIR calls continue. Denying a chart read over licensing would be a patient-safety event, so the system is built so it cannot.

Questions we get on the first call.

Is there a free way to start?

Yes. LangCare is free and open source, and the sandbox runs the console on synthetic patients. Design partners get the full platform in their cloud for their first tenant while we scope the paid rollout together.

Do you charge for the model calls?

No. Inference runs on your provider, in your account, at your rates: Bedrock, Gemini Enterprise, Azure AI Foundry, Databricks Model Serving or the Anthropic API.

What counts as a second tenant?

A second hospital or health system with its own EMR connection and entitlement. More agents, users or calls on an existing tenant are included.

Do we need a BAA with Kartha?

By default, no. PHI never reaches Kartha because the platform runs in your account with your keys. Your BAA is with the hospital, as it is today.

Can we author our own skills?

Yes, and they are never gated. Customer-authored skills run in every licensing state, including grace.