I’m genuinely confused about how usage pools are handled on the Team plan compared to Personal Pro. Why do Team seats lose the dedicated, very large “auto” model pool, meaning our massive daily cache reads (from large codebases) directly drain our paid credits? What is the recommended workaround here, or does it make more financial sense to ask my company to cancel my Team seat and just reimburse me for a Personal Pro+ account instead?
Hey, this is a structural difference, not a bug. On Pro you have two separate pools: $20 of API usage for specific models like Claude 4.6 Sonnet, GPT-5, etc, plus a separate generous Auto + Composer pool on top. On the Team plan there’s one shared $20 pool for everything, Auto, Composer, and specific models all pull from the same $20.
There’s also a pricing detail: on the Team plan, all non-Auto requests include a Cursor Token Fee of $0.25 per million tokens on top of the model’s API price. On Pro there’s no such fee, the only extra is a 20% upcharge on Max Mode.
On your question about Pro+ for a solo dev: if you don’t need team features like SSO, centralized billing, admin dashboard, or usage analytics, Pro or Pro+ gives you more capacity for less money. More details and docs:
- Full answer in a similar thread: Clarification on Cursor Pro vs Team Plan – Agent Usage and Token Allocation - #7 by deanrie
- Models & Pricing: Models & Pricing | Cursor Docs
Within the Team plan, there’s basically one workaround: use Auto where it makes sense since it’s cheaper and doesn’t have the Token Fee, and save specific models for tasks where you really need them. About cache: if you’re seeing a huge volume of cache reads from a large codebase, you can reduce it using .cursorignore and tighter context so it doesn’t index or pull everything.
What does “Composer” mean in this context? Any Composer* model?
Yes, any Composer model, Composer 2.5 and earlier versions.