Control AI access
Last validated:
Aperture uses Tailscale's identity layer to automatically authenticate users. Control which users can reach the Aperture instance through tailnet access control rules, then use Aperture grants to define which models each user or group can access. To reference Tailscale groups in grants, enable visible groups. Aperture is deny-by-default: without grants, users can connect but cannot access any models.
For background on how identity works, refer to how Aperture works. To understand the grants model itself, including deny-by-default access and precedence, refer to how Aperture grants work.
How Aperture grants work
Understand the Aperture grants model, including deny-by-default access, additive allow-only rules, capability matching, and how Aperture resolves precedence when multiple grants apply.
Grant access to models
Configure Aperture grants to control which models each user or group can access.
Grant access to MCP tools
Configure Aperture grants to control which MCP tools, resources, and templates users can access.
MCP server proxying
Aggregate tools and resources from remote MCP servers through Aperture with identity-based access control.
Set up admin access
Configure administrator roles for managing Aperture settings and accessing all user data.
Visible groups for Aperture
Use Tailscale groups and SCIM groups to control access to Aperture models and admin roles.
Use guardrails to enforce policies on individual requests, such as scrubbing PII or blocking requests that violate compliance rules.
Guardrails
Inspect, modify, or block LLM requests before they reach the provider.
Set up a guardrail
Configure a pre-request hook to inspect, modify, or block LLM requests before they reach the provider.