Skip to main content

Billing & Budgets

Caged charges for two things: compute time and LLM tokens (when using Caged-provided models).

Pricing

Compute

Paused sandboxes incur no compute charges.

Plans

Overages beyond included compute are billed at the standard rates.

Budget Guards

Set a maximum dollar amount per sandbox. When the budget is reached, the sandbox is automatically destroyed.

Via CLI

Via Config

Via API

Budget Alerts

Automatic alerts fire at budget thresholds: Customize thresholds in Dashboard → Settings → Alert Rules.

Monitoring Spend

Real-Time

Dashboard

The dashboard shows:
  • Per-sandbox cost breakdown
  • Daily/weekly/monthly spend charts
  • Cost by agent type
  • Projected monthly spend

API

Cost Optimization Tips

  1. Use idle timeouts — Auto-pause inactive sandboxes to stop compute charges
  2. Set budgets — Always set a budget guard to prevent runaway costs
  3. Right-size resources — Use 1 vCPU / 512MB for simple tasks
  4. Use snapshots — Pause and snapshot instead of keeping sandboxes running
  5. Clean up — Destroy sandboxes you’re done with