Reduce AI Agent Costs and Mistakes with Semantic Tool Selection

How to Add Your AWS Promotional Code

Step-by-step: redeem the credit code in your AWS account

Presentation Deck EN (PDF)

English slides from the talk

Presentation Deck ES (PDF)

Spanish slides from the talk

Reduce Agent Errors and Token Costs with Semantic Tool Selection

Blog post with detailed walkthrough and code examples

Reduce errores y costos de tokens en agentes con seleccion semantica

Blog post en espanol

Demo: Semantic Tool Selection

Complete 29-tool travel agent with FAISS filtering

If you find it useful, give it a ⭐!

Alternative: Tool Selection with a Decision Model

One hook (BeforeModelCallEvent) swaps in the relevant tools before the model runs; the selector is pluggable: vector search or a decision model (Jev, Laya). Measured over 12 queries: send-all 10/12 at 55,924 chat tokens, hook+vector 11/12 at 19,605 chat + 391 embed, hook+decision 11/12 at 16,333 chat + 0 embed.

If you find it useful, give it a ⭐!

Strands Agents SDK

Framework for building production-ready multi-agent systems

AgentCore Gateway: semantic search for tools

Production-managed semantic tool routing over MCP (x_amz_bedrock_agentcore_search)

Internal Representations as Indicators of Hallucinations in Agent Tool Selection

Research: how LLMs internally represent tool choices

Solving Context Window Overflow in AI Agents

Research: impact of context size on tool selection accuracy