MCP vs. Function Calling vs. A2A
NewThree ways AI systems connect to other things, often listed together under 'agents'. Function calling (tool use) is a model's built-in ability to ask an application to run a specific function. MCP (Model Context Protocol, from Anthropic in 2024) is a shared standard for connecting AI apps to tools and data, so a connection built once works across many apps. A2A (Agent2Agent, from Google in 2025) is a standard for separate AI agents to talk to each other. Each is a level up from the last.
How to tell them apart on a resume
Function calling / tool use
Tool definitions, JSON schemas, structured output, OpenAI or Anthropic API tool use — one application giving its model a set of actions.
MCP
Built MCP servers, connectors to internal systems, Claude Desktop or IDE integrations, reusable tools across AI apps.
A2A
Multi-agent systems, agents from different vendors or teams working together, agent cards, orchestration across organisations — still early.
The question that settles it
“When your AI used a tool, was that tool defined inside your own app, exposed through an MCP server others could reuse, or provided by a separate agent?”
Read the full definitions
Open the full tool for the other look-alike pairs, role profiles, and the JD decoder.