Metered at the level of execution, not by seat. You see what each application, team and run actually costs.
Author an agent and run it on your own machine. No card required.
For teams putting an agent in front of real users.
For teams running agents at scale, or publishing capabilities to others.
Dedicated environments with the controls your security team requires.
Billed monthly, cancel any time. Usage beyond the included amount is charged at the rates below. Self-hosting the runtime is free on every plan.
Model and tool consumption is captured and persisted per agent, at the level of execution. You see what each agent costs before the invoice does.
| What is metered | What it covers | Rate |
|---|---|---|
| Model tokens | Inference across every provider the intelligence plane routes to. | Provider cost + 10% |
| Runtime hours | Active runtime for long-running, scheduled and always-on agents. | $0.08 per agent-hour |
| Tool and MCP calls | Invocations through attached tool servers and integrations. | $2 per 10,000 calls |
| Memory and storage | Durable knowledge, tool-output persistence and trace retention. | $0.20 per GB / month |
| Marketplace revenue share | Payouts to developers, strategists and capability owners. | 20% Maturing |
An always-on agent costs roughly $58 a month in runtime, before model tokens. Bring your own model keys on any plan and you pay the provider directly, with no platform margin.
No. The runtime is self-hostable and runs against your own database. Managed services handle specification drafting, the capability catalog and metering.
Yes. Direct mode uses your keys; managed mode lets Kodeus handle routing, rate limits and fallback.
They are enabled per deployment and metered by the work they do, so teams that do not use them never pay for them.
Author a spec locally with the open SDK. Move to managed operation and governance when the application is real.
