---
title: "Discussion on \"Your agent shouldn't know about tools it can't use\n\" | Hashnode"
description: "Discussion on \"Your agent shouldn't know about tools it can't use\n\". Use AgentCore Gateway, Cedar and Strands so agents only see tools they may call. Search results aren’t policy-filtered, so intersect them."
url: "https://hashnode.com/posts/your-agent-shouldn-t-know-about-tools-it-can-t-use/6acafa2b6bbff32997081d4d"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Your agent shouldn't know about tools it can't use](https://blog.gainesai.com/your-agent-shouldn-t-know-about-tools-it-can-t-use?utm_source=hashnode&utm_medium=feed)

Use AgentCore Gateway, Cedar and Strands so agents only see tools they may call. Search results aren’t policy-filtered, so intersect them.

Oct 11 · 16 min read · blog.gainesai.com  
Read the article

-   [#aws](https://hashnode.com/tag/aws)
-   [#cedar](https://hashnode.com/tag/cedar)
-   [#security](https://hashnode.com/tag/security)
-   [#amazon-bedrock](https://hashnode.com/tag/amazon-bedrock)
-   [#ai-agents](https://hashnode.com/tag/ai-agents)

## Written by

[James Gaines](https://hashnode.com/@gainesai)

@gainesai · Morgantown, USA

  
[View profile](https://hashnode.com/@gainesai)

Building and breaking things in the cloud

AWS cloud architect and AI enthusiast based in Morgantown, WV. I write about building on AWS, exploring AI agents and automation, and the occasional deep dive into cloud security. Follow along as I share what I'm learning, building, and breaking in the cloud

## Responses 3

Markdown supported · ⌘↵ to post

[Kartik N V J K](https://hashnode.com/@kartik-nvj)

[Oct 12](https://hashnode.com/posts/your-agent-shouldn-t-know-about-tools-it-can-t-use/6acafa2b6bbff32997081d4d/comment/6acd2c77e731423e382c8bcc)

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?

[Puneet Khandelwal](https://hashnode.com/@lifestyle4u)

[Oct 12](https://hashnode.com/posts/your-agent-shouldn-t-know-about-tools-it-can-t-use/6acafa2b6bbff32997081d4d/comment/6accba3ea4c981036e5cc617)

So much cleaner for the model's context window. Hiding unavailable tools entirely stops agents from looping on dead ends.

[Salman Parvez | ML Systems](https://hashnode.com/@ml-systems)

[Oct 11](https://hashnode.com/posts/your-agent-shouldn-t-know-about-tools-it-can-t-use/6acafa2b6bbff32997081d4d/comment/6acba2f9855794dc4f734697)

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.
