Opening soon
MCP (Model Context Protocol)
Learn how the Model Context Protocol works and build an MCP server with tools and resources.
5 chapters · 22 lessons
Want a note when it opens?
Join the waitlist and we'll email you when the course is ready.
What you'll cover
The chapters
The plan for the full course. Lessons open when the course does.
- 01Why AI integrations need a shared protocol4 lessons
- 02How MCP hosts, clients and servers work5 lessons
- 03Building and testing your first MCP server4 lessons
- 04Designing tools and their input schemas5 lessons
- 05Serving data with MCP resources4 lessons
Self-check
How much do you already know?
5 quick questions. Pick an answer to see why it's right or wrong. Nothing is saved.
Question 1 of 5
Which field is present in a request but absent in a notification?
Question 2 of 5
According to JSON-RPC 2.0, which error code range is reserved for server-defined errors?
Question 3 of 5
Can a JSON-RPC response message contain both 'result' and 'error' fields at the same time?
Question 4 of 5
In JSON-RPC, what must be true about the `id` in a response compared to the originating request?
Question 5 of 5
In MCP architecture, what is the primary responsibility of the Host?
In the news
Recent stories on this topic
From NewsPals, our sister newsroom.
- Keysight Advanced Design System 2027 Turns Agentic AI Into an EDA Shift (opens in new tab)
Keysight is making RF agents use macros, code, and MCP servers instead of asking LLMs to guess their way through schematics.
- X’s Ads Model Context Protocol server brings AI agents to ad ops (opens in new tab)
X is moving MCP from developer plumbing into advertiser workflows, where third-party AI agents may create and edit campaigns.
- X Ads MCP Opens Campaign Operations to Third Party AI Agents (opens in new tab)
The launch moves ad creation and editing into AI tools, which makes permissions, logs, and approvals less optional.
- Grafana gcx and MCP Server GA says agents need governed telemetry, not dashboard chat (opens in new tab)
The useful part is not a chattier dashboard. It is giving AI workflows controlled access to the facts production already knows.
- Expedia Just Opened Its Platform to AI Agents. Here Is What Every Product Builder Needs to Know. (opens in new tab)
Expedia's move to let partner AI agents act inside its platform is a masterclass in how platforms must redesign themselves for a world where the user is not always human.
Key terms
Words you'll meet in this course
- MCP server
- A program that exposes tools, resources or prompts to AI applications through the Model Context Protocol.
- MCP host
- The application a user works in, such as an AI desktop app or code editor, that connects to MCP servers through clients.
- JSON-RPC 2.0
- A lightweight format for remote procedure calls sent as JSON; MCP uses it for every request, response and notification.
- Capability negotiation
- The opening handshake in which an MCP client and server tell each other which protocol features they support.
- JSON Schema
- A standard way to describe the shape of JSON data; MCP tools use it to declare the arguments they accept.
- stdio transport
- Running an MCP server as a local process and exchanging messages over its standard input and output streams.
- Resource URI
- An identifier, such as file:///logs/app.log, that names a piece of data an MCP server can return to a client.
Help us choose
Which course should open next?
Pick the one you'd take first. Your vote helps us decide what to finish next.