Cut AI token usage by 96%? Here’s how AWS Strands Agents does it.
from The New Stack Podcast Hosted by Amazon, AWS, Frederic Lardinois, The new Stack, Morgan Willis
In this episode of The New Stack Makers, AWS developer advocate Morgan Willis demonstrates Strands Agents, an open source agentic framework with rapid adoption since its launch. Using a simple accounting API, she walks through three approaches to retrieving a customer’s latest invoice, highlighting how design choices dramatically impact efficiency. The initial method maps each API endpoint to a separate tool, requiring five chained calls and consuming about 52,000 tokens. By shifting to intent-based tools—focused on outcomes rather than individual data operations—the same task is completed in a single call using just 2,000 tokens, improving both efficiency and reasoning. In a third iteration, tools are hosted on a remote MCP server via AWS Agent Core Gateway, with semantic search limiting the agent’s toolset to only what’s relevant per query, further reducing token usage. Willis emphasizes that narrowly scoped agents outperform general-purpose ones, delivering better speed, accuracy, and context efficiency. Designing smaller, specialized agents with tailored tools is key as tool ecosystems expand. Learn more from The New Stack around the latest with Strands and MCP: AWS Launches Its Take on an Open Source AI Agents SDK What Is MCP? Game Changer or Just More Hype? MCP’s biggest growing pains for production use will soon be solved Join our community of newsletter subscribers to stay on top of the news and at the top of your game.
Embed this episode
What this episode covers
In this episode of The New Stack Makers, AWS developer advocate Morgan Willis demonstrates Strands Agents, an open source agentic framework with rapid adoption since its launch. Using a simple accounting API, she walks through three approaches to retrieving a customer’s latest invoice, highlighting how design choices dramatically impact efficiency. The initial method maps each API endpoint to a separate tool, requiring five chained calls and consuming about 52,000 tokens. By shifting to intent-based tools—focused on outcomes rather than individual data operations—the same task is completed in a single call using just 2,000 tokens, improving both efficiency and reasoning.
No transcript yet.