So much cleaner for the model's context window. Hiding unavailable tools entirely stops agents from looping on dead ends.
Treating search as a ranker and not an authorization layer is a good catch. What the model sees shapes what it tries, so I'd rather it never see a tool it can't use. Thanks for including what broke along the way.
Kartik N V J K
AI Developer | Making AI reliable, trustworthy & accessible to everyone | Active community contributor
That gap between the Cedar policy allowing 2 tools and semantic search still returning all 6 with full schemas is a real hole, because the model is now reasoning about capabilities it can never call. Even when the gateway blocks the invoke, you have paid for the tokens and widened the surface for a confused-deputy prompt. I would scope the tool-search index per identity, but I am curious whether AgentCore makes you choose between one shared index and per-user latency. How did you end up handling it?