Principles of Building AI Agents (ft Sam Bhagwat)

Join the Tool Use Discord: https://discord.gg/PnEGyXpjaX

Building AI agents is the current trend, but are you doing it correctly? Sam Bhagwat, founder of Mastra, joins Tool Use to breakdown the principles of building reliable AI agents. We discuss why you should use a TypeScript agent framework like Mastra, how to engineer production-grade prompts, and the best patterns for tool calling. Sam explains advanced concepts like observational memory, dynamic agents, and the "lethal trifecta" of security risks. Whether you are building coding agents or enterprise workflows, this episode covers the essential primitives for success.

Mastra: https://mastra.ai/

Get the Book: https://mastra.ai/book

Connect with us

https://x.com/ToolUsePodcast

https://x.com/MikeBirdTech

https://x.com/calcsam

00:00:00 - Intro

00:01:19 - When to use an AI Agent Framework

00:06:05 - Principles for designing Agent Tools

00:11:01 - Human-in-the-loop Architectures

00:17:19 - Working Memory vs Long-term Memory

00:32:03 - The "Lethal Trifecta" of Security Risks

Subscribe for more insights on AI tools, productivity, and AI Agents.

Tool Use is a weekly conversation with the top AI experts.