Starter
$0.00/ month + usagePersistent projects, parallel agents, and full visibility.
- 2 concurrent jobs
- 30 minutes per execution
- 1 GiB storage
- $0.00 monthly credit
- 30 days of detailed history
Pay for the work.
Keep what matters.
Choose a plan for your organization. Fund usage with prepaid credits, and bring your own model key when you prefer.
Pro
$29.00/ month + usagePersistent projects, parallel agents, and full visibility.
- 10 concurrent jobs
- 60 minutes per execution
- 10 GiB storage
- $10.00 monthly credit
- 90 days of detailed history
Scale
$199.00/ month + usagePersistent projects, parallel agents, and full visibility.
- 50 concurrent jobs
- 120 minutes per execution
- 50 GiB storage
- $50.00 monthly credit
- 90 days of detailed history
Every plan includes a 24-hour queue deadline, adjustable downward per request. Limits are shared across your organization. Concurrency is a maximum, not a reservation or immediate-start guarantee.
Usage, in plain language
| Resource | Customer rate |
|---|---|
| Standard cloud execution | $0.008 per minute, prorated per second |
| Managed inference | Published model rate card, including platform markup |
| Your model API key | $0 platform model charge; your provider bills you directly |
| Managed web search | $0.006 per billable search |
| Extra retained storage | $0.10 per GiB per 30-day month, only when enabled |
Model rates are the final retail rates. Cache reads use the input rate; Anthropic cache writes use twice that rate. Output includes reported reasoning tokens.
Model prices are listed in the authenticated model catalog before a run is submitted. Compute and authorized tools still use platform credits with BYOK. Failed work can incur consumed usage; a budget is an admission and gateway limit, not a promise that failures are free.
Purchased credits do not expire. Included subscription credits expire at the paid period’s end. Storage overage requires its own monthly budget. Projects can be exported or scheduled for deletion; detailed run retention is separate from the persistent native conversation state.
Read about persistence and retention