-
Book Overview & Buying
-
Table Of Contents
Agentic Architectural Patterns for Building Multi-Agent Systems
By :
As systems evolve beyond standaloneidx_f1c91d8a prompts toward the agentic architectures described previously, the ability for models and agents to interact reliably with tools and each other becomes paramount.
This involves understanding and potentially implementing key layers of the emerging AI interoperability stack: function calling, the MCP, and the A2A protocol.
Function calling enables LLMs within an agent's reasoning component to intelligently trigger specific tools (e.g., book_flight(destination="Tokyo") in a travel assistant, get_credit_score for a loan application, or execute a local Python script for data analysis). MCP provides a standardized way to describe, discover, and securely invoke tools (including weather services, calculators, vector search utilities, or enterprise-specific APIs such as verify_property_appraisal for real estate applications) as independent, interoperable services, enhancing modularity
A2A offers...