<Section title="Where OAuth scopes show up in the material">## How OAuth scopes are used hereIn this channel's material, OAuth scopes appear in the context of agent security. Alex Rattray argues that limiting what an MCP server exposes is the wrong security lever, because anything in the underlying API remains reachable under the hood — so security has to be enforced at the API layer itself, using OAuth with granular permissions and proper scopes.### Why scopes, not the MCP layer- The MCP layer is just a translation surface; the real authority over what an agent can do lives with the API provider.- Granular permissions mean an agent can be granted narrow rights — read, refund, create — rather than blanket access.- The API layer can verify and enforce those rights on every request, no matter which client or tool surface is calling.<MomentLink seconds="2454"/>[^913c216d-5ccb-51d9-8f42-8feb7c017d9d@2454]### The broader security pictureThe same interview pairs scopes with another defense: one API provider per sandbox, locking network access so nothing can connect out except the target API — api.stripe.com and nothing else. Together, OAuth scopes at the API layer and network-locked sandboxes form the security model for agents that write and execute code.### Key takeawayScopes are the concrete mechanism by which an agent gets least-privilege access to an API — decided at the API layer, not at the protocol or tool layer.</Section>