Agents Integration Patterns
A catalog of named integration patterns for multi-agent AI systems, written to give architects a shared vocabulary the way Enterprise Integration Patterns did for distributed systems.
This repository catalogs integration patterns for systems of autonomous AI agents that collaborate over protocols such as the Model Context Protocol and Agent-to-Agent, orchestrated by frameworks like LangGraph, AutoGen, CrewAI and Spring AI. It aims to give teams a stable, reusable vocabulary for connecting agents to harnesses, models and each other, regardless of framework or protocol.
It draws on Enterprise Integration Patterns (Hohpe and Woolf, 2003), the 12-Factor Agents guidelines, Google's A2A protocol specification, Anthropic's MCP specification, and academic surveys of multi-agent LLM systems from 2025 and 2026.
The catalog opens by describing the protocol foundation — MCP for an agent's downward connections to tools and resources, covering tool invocation, resource access, prompt templates and sampling — before presenting the named patterns themselves.