Build AI Agents with MCP - Model Context Protocol - TS/PY

Explore the basics of the Model Context protocol (MCP), learn about MCP hosts, and install and use your first MCP servers.

Explains how MCP, or model context protocol, standardizes AI actions via tools. Highlights parallels with HTTP, enabling tools like weather or mail readers to extend capabilities beyond chat.

Explore the MCP architecture by identifying the host, client, and server roles, and learn how a host coordinates multiple clients and servers, with servers providing context and handling tool calls.

Discover how MCP hosts enable tools, resources, and prompts, with practical setup in VSCode Copilot and Claude Desktop, plus Cloud Desktop, and straightforward MCP JSON configuration.

Discover where to get MCP servers—from the official MCP catalog on GitHub to VS Code extensions—while evaluating safety and verifying repositories and stars before installation.

Install your first MCP by adding the MCP compass to your MCP host, ensure Node.js and NTP are installed, and test via VS Code or cloud configuration.

learn to install and configure a complex Spotify MCP, including client id and secret setup, credentials, and running it locally or in cloud with proper path configuration.

Engage in an interactive section coding in TypeScript and Python to build and inspect basic MCP servers, including a quote service in VS Code or cloud desktop with Copilot.

Explore building MCP servers or clients using TypeScript or Python, the two popular languages with strong SDKs, and pursue language-specific lessons and certification.

Initialize an npm project for the MCP with TypeScript, creating package.json and a gitignore. Install the MCP SDK, Zod, TypeScript, and TSX, then test with a sample TS file.

Build a hello world MCP in TypeScript using a quote service with happy, set, and neutral quotes, then register a tool to return quotes and test with TSX.

Extend the MCP by registering new tools, including a quotes tool that reads a count from a Zod schema, fetches quotes, handles errors, and returns a content array of quotes.

Initialize a Python project for MCP using the Python SDK, install mcp cli with uv, and create a virtual environment. Confirm MCP development tools are installed.

Learn to build a first MCP in python by creating a hello world server, decorating tools with MCP tool, and testing the server via uv in vscode.

Extend the MCP with a Python tool that fetches a specified number of quotes, handles validation and errors, and exposes a robust, reusable interface.

Learn to use the MCP inspector to test and debug model context protocols in node or python projects, configure transport types, run tools, and handle parameters in real time.

Explore two important MCP concepts—the transport and the primitives—and expand your knowledge through the practical project two Transport Primitives, choosing TypeScript or Python.

Explore MCP transport as the client‑server communication and authentication layer, comparing standard IO and streamable HTTP transports, and learn HTTP server setup via a URL.

Initialize a new MCP project for both Python and TypeScript, create a readme and two core files for MCP logic and the client, and run with the standard IO transport.

Configure a TypeScript MCP server to run over http using express, creating a post endpoint and streamable transport, then start the server on port 4000.

Discover how to enable the MCP HTTP transport in Python with a single-line change, switching from standard IO to HTTP and running an HTTP server via an MCP endpoint.

Explore the MCP primitives—tools, resources, and prompts—and how the application controls files, databases, and APIs to extend the LLM, with tools used 95% of the time.

Discover MCP resources and how they add structured context to AI by exposing data from files, APIs, or databases using resource templates.

Add static and dynamic resources to a TypeScript MCP project, returning file contents and item prices via resource templates and parameterized URIs for the LM prompt.

Learn to build static and dynamic MCP resources in Python by reading a logs text file and retrieving an item price by id via an MCP resource.

Explore the difference between resources and tools in MCP, and learn to optimize prompts by using a dedicated tool that returns only error lines from logs.

Explore MCP prompts, the final primitive in the model context protocol, and how prompts enable user-controlled, structured messages to language models.

Demonstrates MCP prompts in action with Python and TypeScript, showing a function that accepts topic and paragraphs to generate a history report. Covers VSCode and Copilot prompt usage.

Explore building an MCP-based application—a wrapper over an online store—using TypeScript or Python, order pizzas with an AI client, and check order status.

Clone or download the node project repo, run npm install, then npm start to run the server and access the UI at localhost:3000 for MCP testing.

Learn how websites use http calls and fetch requests, including url, method, headers, payload, and responses, and apply this with the MCP client in TypeScript or Python.

Start implementing our pizza MCP with TypeScript and build the project now, using the finished code available in the course resources.

Begin a TypeScript and Node.js project, build an API client to fetch the daily menu, and create an MCP tool returning menu entries.

Learn to display images in MCP tool responses by fetching image URLs and converting them to base64. Embed the image payload with mime type alongside the menu text.

Build an MCP make order tool in TypeScript that posts an order to a pizza site using input and output schemas with Zod to capture sender and items.

Develop a check order status tool within the MCP framework by building a get request to /orders, extracting the status from the response, and testing the workflow.

Begin building the awesome pizza MCP server in Python, using the course’s ready-to-use code in the Python project resources. Access the complete implementation as you follow along.

Begin a Python MCP starting project by installing uv, adding the mcp cli, and creating separate MCP server and execution modules; test via main standard io.

Create a daily pizza menu tool by making an HTTP request to the backend, using a Python API client with requests, and exposing it as an MCP tool.

Build a pizza menu with images for the MCP server by fetching image URLs, encoding images to base64, and returning text and image content in a call tool result.

Develop a complex pizza order tool with Python types and MCP. Define order items and order details, and expose a make order function that returns an order id.

Build an MCP tool to check an order status by issuing a get request to backend API orders/{order_id} and parsing response data.status, with error handling.

Test the complete pizza MCP workflow by running the site, placing orders, and checking status with AI tools like GitHub Copilot, illustrating a web service as an AI interface.

Gain practical insight into MCP authorization by exploring web services and HTTP authorization with an OAuth-based service, and learn how an MCP server is authorized.

Explore how MCP authorization works by using a dedicated authorization server to issue tokens, handle 401 challenges with the WWW-Authenticate header, and secure MCP requests with OAuth-based flows.

Demonstrates a sample OAuth server to illustrate MCP authorization by starting an Express app with npm, exposing authorize, token, and introspection endpoints, and illustrating token flow and validation.

Explore MCP authorization, implement a protected MCP route with authentication middleware, and verify bearer tokens via the OAuth server's introspect endpoint and well-known endpoints in TypeScript or Python.

Master testing oauth applications in vscode by understanding the auth flow—from MCP post to 401 to the well-known oauth protected resource—and clearing cached state by removing dynamic providers and reloading.

initialize a sample MCP project with TypeScript and express, install dependencies, create a server.ts using a streamable http transport, and register a hello tool for an express server.

Implement authentication for the MCP TypeScript server with an express middleware that handles public routes and bearer tokens, then validate tokens via the OAuth server's introspect endpoint.

Initialize a Python MCP project with a fast API library, use the MCP CLI and uv run to serve a hello MCP tool on localhost:8000, with authentication covered next lecture.

Implement python authentication in the MCP project by adding a FastAPI authorization middleware, validating bearer tokens via the introspect endpoint, and enabling CORS for browser requests.

Celebrate finishing the MCP course and your instructor's guidance, and prepare to write and develop your own MCP servers; discover links to other courses and consider leaving a review.