Skip to main content
Your credential selects the standard billing lane. The transport alone does not. MCP is for interactive work inside an assistant. Automated or machine-scale traffic must use the REST API.
Contracted partner accounts can have plan-pool traffic billed to the API wallet under their agreement. account_status does not show which lane applies; confirm it with your StefanBrain contact.

​
REST API rates

The wallet bills per token. One rate applies to every request; there is no model parameter.
  • Repeated tokens in a conversation bill at the cached-input rate automatically.
  • Cache writes bill as ordinary input. There is no separate write premium.
  • An Agent Run that stops at the runtime’s step limit (last_error: "budget_limited") is not billed.
  • Non-token usage, including image generation and web search, bills per use. The wallet history shows the exact charge.
  • Video generation remains on its own per-minute meter.
  • Wallet-billed calls and plan-pool calls remain separate. The credential and any partner billing rule select the lane for each request.

​
Included credit

Your plan adds API credit to the wallet at the start of its billing cycle. Monthly plan credit resets each cycle and is spent before purchased top-ups. Purchased top-ups remain available until spent, for a maximum of 12 months.

​
When the wallet runs low

Each wallet-billed request that starts work first places a temporary hold on the wallet (currently $4.38). A sync tool call releases it when it returns. An Agent Run, a parked ask_stefanbrain, or a submitted job keeps it until that work ends, for at most 8 hours. A request returns 429 api_wallet_exhausted when the balance minus active holds is below one hold, so a wallet can refuse new work before $0 and each concurrent run uses one hold. The response includes wallet.balance_cents, wallet.included_monthly_credit_cents, and wallet.resets_at. Responses that start billable work also return the current wallet state in these headers:
Requests billed to the plan pool, such as stefan_oat_... requests on a standard member account, do not include these headers. Error responses do not include them either. For standard member accounts, plan-pool traffic is not affected by an empty API wallet. Wallet-routed partner traffic remains subject to the wallet. See Rate limits for throughput limits, which are separate from wallet and key-budget limits.