Managed models, BYOK, and spending caps

Every agent needs at least one usable model provider. Choose managed credentials for simpler setup or bring your own key for a direct provider relationship.

Managed credentials

At launch, Managed provisions an agent-specific OpenRouter key. Pneum.ai meters its provider activity and presents the resulting managed-model usage in Billing. The key is not displayed in the browser after provisioning.

The Cloud Console’s billing records do not store prompts. The selected upstream model provider still processes requests under its own service terms.

BYOK

BYOK leaves model setup to the agent workspace. During onboarding or in Profiles Studio, add a supported provider and its API key. The agent validates the credential before treating it as ready for selection.

With BYOK:

• the provider bills you directly;
• the spend does not appear as Pneum.ai managed-model usage;
• provider limits and retention settings belong to that provider account;
• you are responsible for rotation and revocation.

Never paste an API key into chat, notes, a system instruction, or a tool argument. Use the dedicated credential input.

Monthly spending cap

Open Cloud Console → Billing → Managed model usage and edit Monthly spending cap. Enable the cap, enter the amount, and save. The panel shows the current cap, remaining amount, and reset basis.

A cap protects future managed requests after the recorded usage reaches the limit. Export or provider delays can affect how quickly activity appears, so do not treat it as a real-time infrastructure shutdown control.

Removing the cap makes managed model spend unlimited within other account and provider controls.

OpenRouter attribution

When enabled at launch, attribution adds the Pneum.ai application identity to managed OpenRouter requests. It does not send a transcript in attribution headers.

For model selection inside the agent, continue to profiles and models.