Block II · The apps, at a professional level · Lesson 14

MCP, A2A and Agent Skills: the standards that connect AI

You'll learn what an MCP server is and what it's for, how remote servers and MCP Apps with an interface work, how agents talk to each other with A2A and how a skill is packaged.

Go to the course

What you'll see in this lesson

  1. MCP: the USB-C of artificial intelligence

    MCP is an open standard for connecting an AI to your tools and data: one connector works with many apps.

  2. Tools, resources and prompts

    An MCP server offers three kinds of pieces, and each is controlled by someone different: the model, the app or you.

  3. Local and remote servers

    A local server runs on your computer and talks over stdio; a remote one lives on the internet and connects over HTTP with a URL.

  4. Remote with OAuth: sign in without your password

    Protected remote servers use OAuth 2.1: you sign in on the service's website and the app receives a token with limited permissions.

  5. MCP Apps: interactive interfaces in the chat

    With the MCP Apps extension, a tool can return a small interactive web page, like a chart, a form or a map.

  6. A2A: when agents talk to each other

    A2A is an open protocol for agents from different companies and programs to hand each other tasks, without revealing how they work inside.

  7. Agent Skills: package your know-how

    A skill is a folder with a SKILL.md file of instructions and, if needed, scripts, references and templates.

  8. Progressive loading: only what's needed

    The agent doesn't read every skill in full: first only its name and description, and the rest only when it needs it.

  9. Summary: the standards that connect AI

    MCP connects agents with tools, A2A connects agents with each other and Agent Skills package how to do each task.

Tools and concepts

Sign in with your Google account to access this lesson and the whole course.

Go to the courseSee the full syllabus

More lessons in this block