UsePod

UsePod is an optional marketplace for model inference. In HivemindOS you can use it as a prepaid model provider, or follow the advanced host path when you deliberately want to offer supported local compute.

Set Up UsePod For An Agent

1. Open the agent’s model or provider settings.
2. Choose UsePod.
3. Complete the guided registration.
4. Fund the displayed deposit address with only the amount you intend to use.
5. Refresh the available models and run the small connection test.
6. Choose a spending preset or set explicit price limits before normal use.

HivemindOS stores the provider setup through the approved credential path and shows the latest available balance and route information when UsePod reports them.

Prepaid And Managed Billing

Two billing paths may be available:

• Your UsePod setup uses the provider account or prepaid balance you configured.
• HivemindOS-managed billing uses eligible HivemindOS service credits and a server-held provider connection.

The current app shows which path the selected agent will use, the applicable limits, and the authoritative price or fee before paid use. Do not rely on an old documentation percentage or a locally edited setting for managed-service billing.

Control Spend

Use the provider presets as a starting point:

• Cheapest favors strict price limits.
• Balanced keeps guarded routing with broader availability.
• Fast allows more expensive routes when speed matters.
• No cap removes the provider price ceiling and should be used deliberately.

Provider price limits are separate from wallet budgets, company budgets, and managed-credit balance. A request must satisfy every control that applies to its billing path.

Offer Local Compute

Hosting is an advanced, separate decision. It can expose an approved local model backend to marketplace work and may require identity keys, a funded operator bond, service supervision, and network configuration.

Start from the Fleet machine’s compute-hosting action. HivemindOS checks for a compatible local model server, discovers models, prepares a managed worker configuration, and refuses to go live when readiness checks fail. Review the advertised models, concurrency, network exposure, expected earnings policy, and stop controls before choosing Go live.

Do not assume that connecting UsePod for inference also enables hosting. The demand and host paths have different authority and risk.

Privacy And Availability

Requests sent to UsePod leave your machine and are handled under the provider’s marketplace and routing model. Keep sensitive local-only work on an appropriate local model.

Marketplace capacity can change. A model that was available earlier may be missing or priced differently later. HivemindOS should report the unavailable route rather than silently switching to a provider with a different billing or privacy boundary unless the selected model route explicitly includes that fallback.

For provider-specific protocol details, use the UsePod documentation.