How are you charging for AI usage: per seat, per token, or per action?

Hey everyone 👋

We’re building billing for AI products, and one thing keeps coming up: flat monthly pricing doesn’t fit AI.

One customer asks 10 questions a month. Another runs 10,000 tasks. The cost to serve them is wildly different, but many teams still charge them the same price.

We’ve seen builders try a few approaches:

1. Per seat: simple, but heavy users cost you money

2. Per token: accurate, but customers don’t understand tokens

3. Per action: one question answered or one task completed, easy to explain but harder to meter

4. Prepaid credits: predictable for you, but customers have to top up

5. Pay after use: frictionless, but there’s a risk of unpaid bills

We built our open-source Billing MCP around per-action billing, with prepaid wallets, post-billing and real-time balance checks, so a customer can’t overspend. But we’d rather learn from you than pitch you.

Tell us:

• How do you charge for AI usage today?

• What’s gone wrong with your pricing model so far?

• What would your ideal AI billing setup look like?

The best answers will shape what we build next, and we’ll reply to every one. 🙌