Does Someone Who Buys Your AI Agent Actually Own Anything?
Three different places an agent can live, and why that's a pricing decision, not just a technical one.
Pricing an agent for The Automated CMO ran straight into a question I hadn't fully answered for myself yet: if someone buys an agent I built, do they actually own anything, or are they renting access to something that keeps living on my servers?
I'd been treating that as one question with one answer. It isn't. The answer depends entirely on where the agent actually lives, and there are three real options, not one.
Here's why this isn't just a technical distinction, and why it needed to become a pricing framework instead of staying an architecture note. Each of the three is a genuinely different product, and they should be priced like different products, not like three tiers of the same subscription.
The three-tier ownership ladder
Skill or config
Runs in the buyer's own Claude account, fully theirs, nothing hosted by you. Lowest friction, lowest price.
Deployable instance
A template the buyer spins up in their own Vercel/database accounts. They own the running app; you designed it. A real product, priced above simple SaaS access.
Full codebase license
Sold outright, no strings attached. Highest ticket, because the buyer is purchasing independence from you entirely.
Skill ownership is the lowest friction to deliver and the lowest price point, because what's being sold is essentially configuration, not infrastructure. A deployable instance is a real product with a bigger price tag than ordinary SaaS access, because the buyer is getting something that runs independently of me, on infrastructure they control, not access they lose if they stop paying. A full license is the highest ticket of the three, because what's actually being purchased at that tier is independence itself: no ongoing relationship to me required, no dependency on anything I keep running.
If you're building and selling AI agents or tools yourself, decide which of these three you're actually offering before you try to price it. The pricing question doesn't resolve cleanly until the ownership question does first.
Go deeper
The Assessment Tool That Almost Only Worked for People Already Using Claude →
How a single question about distribution flipped an entire product's architecture.
The Product Name I Had Locked Turned Out to Belong to Someone Else →
A near-miss with a name collision, and what it taught me about naming as positioning, not branding.
Get new guides and builds in your inbox
No noise. New artifacts, skills, and lessons when they drop — and early access when agents launch.
Want a full structured path instead of one-off guides?
That's Learn AI with AI — a sequenced course that takes you from zero to building real things, with community and accountability built in.
Check out Learn AI with AI →