The Standard Playbook
App stores take 30%. Payment processors take 2.9%. SaaS marketplaces take 20–40%. The justification is always the same: we provide the distribution, you provide the content.
We charge 20% and give builders 80%. Here is the real reason why.
The Quality Flywheel
When a builder earns meaningful money — not symbolic money, meaningful money — from their agent, they invest more in making it excellent. They write better documentation. They tune the system prompt. They add MCP integrations that handle edge cases. They respond to user feedback.
A builder earning $500/month from their agent treats it like a product. A builder earning $50/month treats it like a side project. A builder earning $5 treats it like an experiment.
We are betting on the flywheel: better economics → better agents → more usage → more revenue for builders → even better agents.
The Math
At 80% revenue share:
| Monthly API calls | Price per call | Builder monthly revenue |
|---|---|---|
| 10,000 | $0.01 | $80 |
| 100,000 | $0.01 | $800 |
| 1,000,000 | $0.01 | $8,000 |
An agent doing 100,000 calls per month at $0.01/call earns its builder $800/month. That is enough to justify maintaining and improving it. That is the threshold that matters.
What We Get
Our 20% funds inference costs, infrastructure, customer support, and platform development. At scale, 20% is more than sufficient — the gross margin on AI inference is improving every quarter as model costs fall.
More importantly, we get a marketplace of high-quality agents that users actually want to pay for. That is worth far more than a higher take rate applied to a mediocre catalogue.
The Long Game
We believe the agent marketplace that wins will be the one where builders earn the most. Not the one with the most features, the best UI, or the lowest inference prices. The one where creating excellent agents and publishing them is a viable economic activity.
That is the game we are playing.