Pay for what you use. That's the whole model.
No subscription, no monthly minimum, no charge per user. Buy credit, use CodeCobra, and watch it draw down. If you have a quiet month, you have a small bill.
Rates
Per million tokens, in US dollars. Pick a capability class per task — or leave it on Striker and forget about it.
Hatchling
Straightforward, well-defined work. Simple questions, small field changes, routine edits.
- Input
- $0.30
- Cached input
- $0.06
- Output
- $1.00
Striker
The default, and the right choice for most work. Investigations, reports, building a module, debugging.
- Input
- $2.80
- Cached input
- $0.52
- Output
- $8.80
Apex
The hardest problems. Intricate debugging, large refactors, anything where a wrong answer is expensive.
- Input
- $6.00
- Cached input
- $0.60
- Output
- $30.00
Why the cached rate matters
A long conversation re-reads its own history on every turn, and that history is cached. So the tenth message in a chat costs a fraction of what it would at the full input rate — which is why a thorough investigation usually costs far less than people expect. On Hatchling, cached input is $0.06 per million tokens.
What does that actually come to?
A token is roughly four characters of English. Reading a 50-row spreadsheet is a few thousand input tokens; a detailed development plan is a few thousand output tokens. In practice most questions cost cents and most module builds cost a few dollars.
New accounts start with free credit, so you can find out on your own data before deciding anything.
Teams
Create an organization, invite as many people as you like, and control access with teams and roles. There is no per-seat charge — a team of two and a team of two hundred pay the same rates. Organizations can pay by invoice and see spend broken down by member and by project.