Model and provider access
Allow, block or restrict AI providers, their model categories and individual models to specific teams - while admins keep access to everything your data residency policy permits.
StickyPrompts gives your workspace a wide choice of models from many providers, so people can pick the best model for each job. Model & Provider Access lets you decide how wide that choice is: open everything to everyone, block a provider your company has not approved, or reserve certain models for the teams that need them.
How it works
On the Workspace page, open Model & Provider Access. Under Providers, each provider has its own row with:
- its name and a GLOBAL badge,
- a summary of what is allowed, for example “23 of 23 models allowed across 2 categories”,
- a rule: Allow, Block or Specific teams,
- an arrow to expand the row and set rules for the provider’s categories and individual models.
| Rule | Effect |
|---|---|
| Allow | Everyone in the workspace can use the provider’s models. |
| Block | No one can use them, except workspace admins. |
| Specific teams | Only the teams you choose can use them. |
Restrict a provider
- Open Model & Provider Access
On the Workspace page, select Model & Provider Access.
- Find the provider
Providers are listed one per row, for example StickyPrompts, OpenAI, Anthropic, Google, xAI, Mistral AI and Xiaomi.
- Choose the rule
Open the drop-down on the right of the row and choose Allow, Block or Specific teams. For Specific teams, pick the teams that should keep access. Teams are set up in Users and teams.
- Refine by category or model, if needed
Expand the row to go finer: allow a provider but block one category, or keep a particular model for one team.
Rules that always apply
The order of precedence is simple:
- Data residency decides which models can run at all in your workspace.
- Model & Provider Access decides who, among your members and teams, can use the rest.
- Admins can use everything left after step 1.
Common setups
| Goal | Setup |
|---|---|
| Only approved providers | Set every provider your company has not approved to Block. |
| Pilot a new provider | Set it to Specific teams and pick the pilot team. Widen it to Allow when the pilot is done. |
| Keep some models for one team | Expand the provider and restrict the relevant category or model to that team, for example Marketing. |
| EU-served inference only | Do not do this here - use Data residency. It keeps the rule tied to where models are served, not to who makes them. |
The Models column on Team Management shows which models each team can use.