How to Monetize AI Agents: Lessons from Cursor's Pricing
AI agent monetization phase has come sooner than expected
Each month, this newsletter is read by over 40K+ operators, investors, and tech / product leaders and executives. If you found value in this newsletter, please consider supporting via subscribing or upgrading. This helps support independent AI analysis.
AI agent companies are getting serious about monetization, starting with coding or prototyping AI agents. So it’s worth following their moves to learn how to monetize AI agents, which are notoriously costly to serve.
Recently, Cursor sparked backlash on Reddit and X after quietly changing its pricing for the Pro ($20/mo) tier, and introducing a new Ultra ($200/mo) tier. Some users were hit with surprise $1,000+ bills, which were quickly refunded in a damage control move.
But here’s the thing: Cursor didn’t go rogue.
Since April 2025, every major AI agent company in this space—Vercel’s V0.dev, Replit, Lovable, and others—almost simultaneously tweaked its pricing. Cursor just got unlucky, while others flew under the radar. Btw, this also applies to voice AI agent infra providers as well.
So what’s the new monetization playbook?
Ditch request-based or generation-based pricing, which can cost (the companies) a fortune when AI agents run for a long time
Move to token-based billing tied directly to model inference
Give API credits directly to the users, and have them pay for overages
Remove the whole “fast request”, “slow request” thing, e.g. don’t charge for faster generations. That’s so 2023.
Obfuscate what models are being used by defaulting to private-label models (e.g. “Vercel Small”, “Cursor Auto”, etc.
Shrink or eliminate free tiers
Introduce an expensive top-shelf tier (e.g. “Ultra”, “Max”)
…etc
But why now? I believe the pricing change is a broader reflection of AI agent companies refocusing to serve the enterprise market and go for a larger share of their wallets, instead of competing for users in the hobbyist / startup tiers.
It’s also a sign that the market has matured significantly in the last 6 months, and the second derivative of growth has already started to slow, which means companies need to focus on profitability (below: OpenRouter token volumes growth is slowing, though some of this is seasonality).
So in this post, I’ll provide my thoughts on the recent AI agent pricing trends, why these changes are happening now, and whether these changes are affecting non-coding AI agents as well.
The New Pricing Strategy for Maximal AI Agent Monetization
So what is the new pricing playbook exactly, and why are (some) people upset?
Since May 2025 or so, every popular AI agent company consolidated toward the same pricing strategy, which consists of 5 elements.
Interestingly, many of these strategies are seem to be inspired by Casinos and Slot Machines mechanics, which is essentially vibe programming is (/s).
