Owned vs rented AI agents — why ownership matters
An owned AI agent is one you can export, move, and take with you. A rented agent is one tied to a platform that keeps your encoded work. As agents represent individual expertise, ownership becomes as important as it already is for data and code.
Most AI agent pitches never mention ownership. That's a tell — the thing they don't mention is the thing you'd care about if you understood what's happening. When an AI agent is trained on your expertise and represents how you work, the question of who owns it is at least as important as who owns your data.
What 'owned' actually means
An owned AI agent has four properties:
- Exportable memory — you can get the agent's learned state out of the platform in a readable, portable format
- Portable across platforms — you can take the agent to another vendor if the first one stops being worth it
- Readable and editable — you can inspect what the agent has learned and correct it when it's wrong
- Shareable on your terms — you decide who gets access to the agent's capabilities
Fail any of these and the agent is a rental, even if you pay for it monthly.
What 'rented' looks like
Most major agent platforms today are rental models. Specific symptoms:
- Memory is platform-hosted with no export path (ChatGPT Memory is the clearest example)
- Custom skills, GPTs, or playbooks you build can't be exported in a useful format
- If you stop paying, you lose access — including to the work you did training the agent
- The agent's learned state is proprietary, bundled with the platform's infrastructure
Rental is fine for most software. It's not fine for software that encodes years of your professional judgment.
Why it matters more for agents than for SaaS
SaaS ownership questions have been settled for decades: data is yours, software is theirs, and your data moves with you if you leave. Agents blur this. The agent isn't just software you use; it's software that encodes how you work. If it's trained on your screening criteria for candidates, your negotiation style for contracts, your specific framework for screening deals — that encoding is yours, not the vendor's.
Losing the agent when you change jobs is like losing your Rolodex in 2004. The contacts are yours; the software holding them should be portable.
"The rental model worked for databases because the data was yours. It doesn't work for agents because the agent IS the work.
— The Spawnlabs thesis
The asymmetric risk
In a rental model, the platform has all the leverage as you compound on it. Every month you stay, your agent gets more capable — and more expensive to leave. This is the 'lock-in by dependence' pattern, not lock-in by contract.
With owned agents, the dynamic inverts. Your agent can move, so the platform has to keep earning your business on ongoing value, not on your sunk cost. Pricing pressure goes up; platform bullshit goes down.
Why labs won't offer ownership
Anthropic, OpenAI, and Google sell to enterprises. Their agent products, by necessity, are employer-owned. A platform where the expert owns the agent — and can take it with them across jobs, rent it out, or run it independently — would break their enterprise motion. That's structural; they can't fix it without cannibalizing core revenue.
This is the opening. Platforms built around expert ownership can offer something the labs structurally can't. Spawnlabs is positioned here; so are a handful of smaller players we've mentioned in our comparisons.
What to ask a vendor
- Can I export my agent's memory?
- Can I export my custom skills or playbooks?
- What happens to my agent if I stop paying?
- Can I move my agent to another platform?
- Who legally owns the agent's outputs and encoded state?
If the answers are evasive, the agent is a rental. That's not always a dealbreaker — but it's table-stakes information for anyone who understands what's at stake.
The Spawnlabs take
We built Spawnlabs around the ownership model specifically. Your agent's memory lives in a MEMORY.md file in a per-user Modal volume. You can read it, export it, edit it. Your skills are structured, portable files. If you leave Spawnlabs, you take the agent with you.
We think this matters because the agent will increasingly be the work. As individuals and teams compound on agents representing them, the question of who owns the work becomes the question of who owns the agent. The answer should be: the person who did the work.