Session concurrency
All websites using the same organization share its simultaneous-session allowance. The concurrency endpoint returns an approximate active count, the limit, and an estimated wait. It is an advisory snapshot, not a reservation: another request can use capacity before your start request arrives, and stale session state can affect the estimate. Session creation remains authoritative and can reject a start even after a successful pre-check. A rejected start does not guarantee that a session was created. Check the returned error before retrying: a429 can reflect an organization limit, service capacity, or a spend limit, not just a need to upgrade. Show the reason to the user and avoid retrying in a tight loop. Counts of rejected attempts can include repeated requests from the same person; they are not counts of unique customers or lost revenue.
Custom avatars and creation requests
Each saved, non-deleted custom avatar in your organization occupies a custom avatar slot. Lab and the API check this count before creating another avatar. Finishing generation or ending a conversation does not free a slot. An allowance describes the limit, not the number of custom avatars currently saved. Check actual slot usage separately before concluding how much capacity remains. If creation is blocked because all custom avatar slots are occupied, waiting for a running conversation to finish will not free one. Review unused avatars or contact Anam Support about the effective limit. Check for personas using an avatar before deleting it. The limits and counts measure different things:- Session concurrency: simultaneous sessions using the organization’s session allowance.
- Custom avatar slot usage: saved, non-deleted custom avatars counted by the creation gate, not a count of avatars currently speaking.
- Creation admission: whether a new request passes slot, feature-access, image-generation, and moderation checks. A free slot does not guarantee admission.
- Avatar inventory: the assets visible in a catalog can include stock avatars and other categories, so a catalog total is not automatically custom slot usage.
- Generated-video job concurrency: a separate limit for script-to-video jobs, not an avatar inventory count.

