Скачать или смотреть AI Agents 7 - Model Context Protocol

Description: This lecture (CSE 491/895 at Michigan State University) introduces two critical design patterns for building robust AI agents:

1. Model Context Protocol (MCP): Learn how MCP standardizes communication between large language models and external tools. We cover the client-server structure, how MCP acts as middleware, available public server implementations, benchmarking with MCC-Bench, and hands-on code examples for building MCP servers and clients.

2. Exception Handling: Explore strategies for detecting and managing errors in AI agent workflows. Topics include validating tool outputs, handling API errors, applying timeouts, and using reflection to recover from failures.

By the end, you’ll understand how to integrate standardized protocols and robust error handling into agentic systems to improve reliability, scalability, and adaptability.