Pay for the work, not the seat
Each plan sets a monthly inference allowance. Inference on DeepSeek Flash costs about $0.13 per million tokens, so allowances go a long way. Usage pauses at the cap rather than surprising you.
Free
$0
Loading plans…
How the allowance works
The header meter shows today's spend, month-to-date, and the percentage of the allowance used. At 100% new runs stop until the month rolls over or the plan changes. Cached input is billed at 2% of the miss rate, so a stable system prompt keeps the real cost far below list price.
No card to start
Free tier is a real workspace with an agent and the full meter.
Cancel anytime
Stripe-hosted portal. Downgrade takes effect at the period end.
Bring your own model
Set the model per workspace. Flash for volume, Pro for depth.