Mirantis Agent Platform FAQ#
Answers to common questions about Mirantis Agent Platform licensing and billing.
What counts as an agent-hour?#
One agent using the platform for one hour. For an agent that runs in a sandbox or a managed agent, it is deployed time — an agent running 24/7 is 720 hours a month. For an agent that connects over MCP, it is active use time.
Stop a sandbox you are not using. A stopped sandbox releases its compute and keeps its volume, so restarting it costs you nothing but the restart.
Are model-provider costs included?#
No. You pay your model provider directly. Mirantis Agent Platform meters inference that goes through managed inference so you can see and cap it, but it does not mark up provider costs.
Inference that does not go through the platform is invisible to it. If cost control matters, require managed inference in your org policy rather than allowing provider domains directly. See Spending limits.
What happens when I hit a spending limit?#
Inference requests are refused once the limit is reached, and they stay refused until the period rolls over or the limit is raised. Limits apply per organization, user, agent, or sandbox, and the tightest one wins. See Spending limits.
Can I try it before committing?#
Yes, and you do not need to talk to anyone first. Install the platform on one disposable cloud VM, on a public HTTPS address, launch an agent, and chat with it. See the Quickstart (VM), or On your laptop to run it on your own machine. When you are ready for a real deployment, see Install for production.
How long is the evaluation, and how do I get a license?#
A new install runs as a full evaluation for 30 days from activation, with nothing limited. After that it needs a license key, which an installation owner enters on the License page. Talk to us for a key, or for a longer evaluation. See Activation and license.
What do I get on the platform?#
Everything the documentation describes is part of the platform, not an add-on tier:
- Sandboxes, MCP-connected agents, and managed agents
- Policies and policy bindings, with org policy as a ceiling
- Credential injection at the network boundary
- All connections — Kubernetes clusters, AWS accounts, MCP servers, and any HTTP API
- The full audit trail
- SSO through OIDC
- Spending limits and cost breakdowns
- Multi-tenant administration — organizations, teams, and projects
For licensing terms and commercial questions, talk to us.