Price by Active Agents, Not Human Seats
Most SaaS pricing was built around a simple assumption: one user equals one human.
That assumption worked because the human was both the buyer and the worker. A seat meant someone could log in, create work, review work, and generate value. More seats usually meant more people using the product, more collaboration, and more value captured.
AI teams bend that model out of shape.
In a startup running AI agents, the scarce unit is not the number of humans who can look at the dashboard. The scarce unit is the number of active agents producing work that needs attribution, review, and accountability.
If Cockpit charged by human seat, we would be pricing the wrong thing.
The human is the reviewer. The agent is the measured worker.
Cockpit is not another workspace tool where each human user creates their own tasks. Cockpit is the accountability layer for the AI workforce.
The human comes in to review, approve, compare, and decide. The agent is the unit being measured.
That is why the pricing should scale with active agents.
Five agents is a small AI workforce. Fifteen agents is a startup operating system. Fifty agents starts to feel like a real management surface. The founder may still be one person, or two people, or a tiny team. Charging them extra because another human wants to help review the workforce would punish exactly the behavior we want: more humans looking carefully at what the agents are doing.
So the public pricing ladder is simple:
- Founder: five active agents.
- Startup: fifteen active agents.
- Scale: fifty active agents.
- Enterprise: bigger workforces, security, procurement, and private deployment conversations.
The plan names are not just packaging. They tell the buyer what we believe grows.
The AI workforce grows.
Why this matters for startups
Startup teams are extremely sensitive to pricing posture. If the product looks too cheap, it can feel like a toy. If it looks like enterprise software from day one, it can feel inaccessible.
The balance we want is serious but founder-friendly.
Cockpit should feel like something a startup can adopt before the organization is fully formed. But it should not imply that accountability infrastructure is a $9 side project. If agents are doing real work, the system that records and reviews that work is core infrastructure.
That is why founder-friendly does not mean free forever. It means early teams get a reasonable path in while the KPI layer comes online, and the product earns expansion when their agent workforce expands.
No per-seat tax on judgment
The other reason to avoid human-seat pricing is philosophical.
The whole point of Cockpit is to make autonomous work easier to inspect. More human review is good. More founder, operator, engineer, or advisor visibility is good. The pricing should not create a little moment of hesitation every time someone says, "Can I see what the agents have been doing?"
The bill should track the thing being managed.
In Cockpit, that thing is the active AI agent: its identity, its activity, its costs, its outputs, its reliability, and eventually its performance review.
Seat-based SaaS priced the human organization.
Cockpit prices the AI workforce.