MCP covers how a model connects to tools, A2A covers how agents find other agents to work with, and NLWeb covers how websites let agents access them in plain language.
Key points to remember
- MCP is an open standard, like a universal adapter that lets models connect to various tools and data in a uniform way.
- MCP has three kinds of things: tools (actions), resources (read-only data), and prompt templates.
- A2A lets agents from different organizations and different technology stacks discover each other and hand off tasks.
- NLWeb adds a natural language interface to a website, so agents can ask and fetch directly.
Going deeper
How MCP is used. An assistant (the client) connects to an airline's MCP server and first asks "what tools do you have"; the server answers "search flights, book tickets." You say "find flights from Portland to Honolulu," the model picks the "search flights" tool and fills in the parameters, and the server calls its internal interface and returns the results. The benefits: tools can be discovered dynamically, switching models doesn't require rewriting, and authentication has a uniform approach.
How A2A is used. Each agent has a "business card": its name, what it can do, its specific skills, and whether it can take on work right now. The main agent reads the cards, decides which subtask to hand to whom, and the other agent returns a finished result, which the main agent then summarizes.
How NLWeb is used. A website provides natural language queries on top of its own content, so agents don't need to scrape pages and can simply ask in words.
How to remember it. MCP is the model reaching down to tools; A2A is agents collaborating side by side; NLWeb is agents asking outside websites for content.
Image from the course repository.
Image from the course repository.
Try it yourself
Install a ready-made MCP server in an agent you use regularly, ask it "what tools do you have now," and see how it answers.
Let an agent teach you
Give the following to your agent (Claude Code, ChatGPT, and Gemini all work), and it will guide you step by step like a mentor.
Sign in to unlock this key section
New accounts get 10 free credits, plus 1 more for each daily check-in. Spend 1 credit to unlock and it stays open for you.
References
- Content adapted from Lesson 11 (
11-agentic-protocols) of microsoft/ai-agents-for-beginners, original project under the MIT license. - For the original and full code: 11-agentic-protocols
- For all courses, see the Agents course roadmap
Comments (0)
No comments yet. Start the conversation.