Usage and costs

The model provider you connect owns request pricing, rate limits, quotas, invoices, and promotional free tiers. BIGGI displays estimated session cost when the provider catalog includes pricing, but the provider dashboard is the billing source of truth.

Control spend

1. Set budgets and alerts in the provider's dashboard.
2. Use the least expensive model that can reliably complete the task.
3. Keep prompts and attached context focused.
4. Compact long sessions when earlier tool output is no longer useful.
5. Stop a session that repeats a failed action.
6. Limit parallel sessions when each session uses a hosted model.
7. Use a suitable local model for work that does not need hosted inference.

Configure a session alert

Set BIGGI: Max Cost (biggi-code.new.maxCost) to a whole-dollar amount. BIGGI shows a non-blocking alert when its estimated session cost crosses that value.

The alert does not stop the session and is not a provider spending limit. Set hard budgets with the provider.

Autocomplete costs

Autocomplete has its own provider and model selection. It can create frequent small requests while you type, so check the chosen autocomplete provider separately from the chat model.

Parallel work

Each Agent Manager session has its own model usage. Multi-version worktrees can intentionally send the same task to several models, so estimate the multiplied cost before starting a comparison.

No BIGGI billing

Managed BIGGI credits, passes, automatic top-up, team credit pools, and gateway billing are not available in the current release.