Managed model gateway
Proliferate's managed gateway route — central keys, per-user virtual keys, and how budgets apply.
A gateway is the path an agent's model calls take to a provider. This page covers Proliferate's managed gateway, which members enable per agent through the authentication route. Running your own gateway instead is self-hosted deployment configuration, not an admin setting — see Bring your own gateway below.
| Who | Members enable the managed gateway per agent; admins configure org budgets and caps |
| Applies to | Cloud and Local |
| Requires | Gateway access on the org plan for the managed gateway; a self-hosted add-on for your own gateway |
| Affects | A user (their gateway usage) and the organization (budgets you configure) |
Proliferate's managed gateway
Turn on the gateway route for an agent and it reaches current models with no key to paste or rotate. Under the hood:
- Proliferate holds central provider keys, so no one on the team needs their own.
- Each account gets a per-user virtual key, so usage is attributable to the person who spent it inside a shared organization.
- Provider secrets are not sent to a machine or a sandbox; only the scoped virtual key is.
- Eligible accounts may receive a free signup credit to try it, depending on configuration and a linked GitHub identity.
Any budgets or LLM caps you configure apply to managed-gateway usage; see Usage & limits for how those limits work today. Treat them as the controls to configure, not as an airtight organization-wide cutoff: a personal enrollment path can route around the organization's selection, and end-to-end enforcement of that case is not yet verified.
Bring your own gateway
Running your own model gateway or inference endpoint is a self-hosted deployment task, not an admin setting. See Model gateway.