Plans are per team, not per person — everyone on the same team shares the same limits regardless of who's logged in.
What each limit covers
| Limit | What it counts | Free | Professional | Enterprise |
|---|---|---|---|---|
| Projects | How many projects a team can have at once. | 1 | 5 | Unlimited |
| Active models | How many models a team can have at once — every engine plus EDA nodes, and a model that ended in error still occupies its slot until you discard it. | 1 | 5 | Unlimited |
| Concurrent runs | How many models can be training at the same time. | 1 | 3 | 10 |
| Team members | How many people can belong to the team. | 1 | 5 | 15 |
| API access | Whether programs can read your results and launch runs — see API keys. | — | ✓ | ✓ |
Allocators don't count against Active models or Concurrent runs on their own — they're lightweight and run after a model already has, so they don't compete for the same capacity a training run does.
What happens at the limit
Projects and active models are hard stops: meryn tells you you've hit the plan's limit and offers to upgrade right there, rather than a generic error — creating a new one isn't possible until you either delete something or move to a plan with more room.
Concurrent runs isn't a hard stop — a model that would push you over the limit is simply queued and starts automatically the moment a slot frees up, no action needed from you. Allocators never enter this queue at all: one runs as soon as its model finishes training, without taking a slot.
Team members is checked when the invitation is sent: one that would exceed the limit is blocked with an "Upgrade to add more" error rather than accepted silently. The check counts current members — invitations that are still pending don't count against the limit until they're accepted.