AI Agent Protocols & Standards

Overview

This skill covers the emerging ecosystem of open standards and protocols for AI agents. These specifications define how agents discover capabilities, communicate with each other, make payments, render UI, and are described declaratively.

Protocol Landscape

How They Relate

┌─────────────────────────────────────────────┐
│  Agent Definition (ADL, AGENTS.md)          │
│  "What the agent is and what it can do"     │
├─────────────────────────────────────────────┤
│  Capability Discovery (Agent Skills, MCP)   │
│  "How agents find and load tools/skills"    │
├─────────────────────────────────────────────┤
│  Agent Communication (A2A, ACP)             │
│  "How agents talk to each other"            │
├─────────────────────────────────────────────┤
│  Payments (x402, AP2)                       │
│  "How agents pay for services"              │
├─────────────────────────────────────────────┤
│  UI (MCP Apps)                              │
│  "How agents render interactive interfaces" │
├─────────────────────────────────────────────┤
│  Runtime (cagent)                           │
│  "How agents are orchestrated and executed" │
└─────────────────────────────────────────────┘

Choosing the Right Protocol

• Building tools for agents to call? Use MCP to expose them as tool servers.
• Packaging reusable knowledge/instructions? Use Agent Skills (SKILL.md).
• Orchestrating multiple agents? Use A2A for inter-agent communication (ACP's REST approach merged into A2A).
• Monetizing an API for agent consumption? Use x402 for micropayments.
• Enabling agent-driven purchases? Use AP2 for secure commerce flows.
• Defining agent configurations declaratively? Use ADL for portable blueprints.
• Guiding coding agents in a repo? Add an AGENTS.md file.
• Serving rich UI from an MCP server? Use MCP Apps.
• Running multi-agent systems locally? Use Docker cagent.