MCP Bootcamp: Build, Deploy & Secure Model Context Protocol

Learn how agentic AI behavior works by comparing chat UI flows to API requests, and extend LLMs with tool use and function calling for live data and web search.

Generate a Gemini API key from the AI Studio website, copy it, and add it to your dot env file in Visual Studio Code to enable the tool calling demo.

Validate the Gemini API key, load environment, and interact with a Gemini backend to fetch bitcoin prices via Binance using a defined crypto price tool, illustrating agentic tool calling.

Install Python and Node.js, follow course resources for versions, and set up MCC into LMS with the competitor and cursor tools, using Visual Studio Code for a compatible workflow.

Clone the MCP course repository from GitHub using Visual Studio. Paste the URL in the source control panel, choose a destination, and open the project.

Install and explore uv, a next generation Python package manager that speeds up dependency setup with one click, enabling fast project initialization on macOS and Windows.

Set up your python environment in vscode by creating a terminal, making and activating a virtual environment, then install dependencies from pyproject.toml with uv sync.

Open the first notebook in Visual Studio Code. Install the Python extension and Python Jupyter extension, select the Python kernel, and confirm the environment is up and running.

Explore the problem of exploding tool integration complexity as providers like cloud, cursor, and Slack require custom API calls; MCPs offer a standard solution for model context protocol.

Explore the MCP architecture with a focus on the MCP server. Hosts connect to service providers like perplexity via a standard USB-C style communication channel.

Understand the MCP architecture: hosts, MCP servers, and the MCP client enable tool and data service integration. Learn how the LLM routes tool use, selecting services like perplexity search.

Explore key functionalities of the MCP, including tours for tool exposure to LMS, resources and prompts integration, and routing and sampling concepts for secure, efficient LLM interactions.

Explore tool use in MCP by creating a simple function that runs a shell command, returns its output, and annotate it with the MCP tool after installing the MCP library.

Expose resources to host LM by implementing resource definitions with list resources and read resource functions. Use a URI and markdown mime type to share course resources with LM apps.

Explore MCP prompts, which inject a reusable instruction set into the LLM to deliver expert-level code reviews focused on core logic, with simple bullet-point outputs.

Download the Cloud Desktop client and the Cursor Eye Editor to set up MacBook integration; both offer free accounts with signups, and MCP only works in the desktop client.

Sign up for zapier and integrate its mcp server into our cloud. Create a simple email sending workflow to demonstrate how diverse automations work within a cloud-based mcp setup.

Set up a Zapier MCP server to send emails by configuring an SMTP server and adding a Gmail tool, then save the OAuth server URL for cloud integration.

Integrate Zapier MCP with Claude to send email summaries using Gmail, by editing cloud config, enabling a local MCP proxy, and authenticating via OAuth.

Add Zapier MCP to cursor by configuring a global MCP server via JSON, using the Zapier token, and enable the Gmail send email tool to share course resources.

Learn to use MCP to integrate cloud tools and fallback to the model context protocol reference implementation, called server everything, when Zapier fails, including a Python add-two-numbers calculator.

Explore the main Model Context Protocol MCP servers: official anthropic references, misery community listings, and corsair’s directory, with integration examples for slack and redshift in cloud workflows.

Explore the MCP development libraries and build an MCP server using the Python SDK. Learn about supported languages and docker deployment, and implement a real-time bitcoin price fetcher from Binance.

Set up a bare-bones python MCP server by installing the MCP package with the CLI extra, managing dependencies via pyproject.toml, and running a minimal Binance MCP MVP.

Implement real-time crypto price knowledge in MCP by building a get_price function that calls Binance API, maps symbols, handles errors, returns JSON, and documents behavior with a docstring.

Discover and manage two essential MCP paths—the python file path and the python executables path—across Windows and Mac, using copy, which python, and the built-in MCP debugger.

Explore the MCP inspector to debug and trace your MCP, connect via Python and npm, view tools and health checks, and interpret red debug output as non-errors.

Explore MCP transport types, including local stdio and remote streamable http, ssh, and server-sent events, and learn when to deploy with stdio for local MVPs versus http-based remote endpoints.

Integrate our own mcp with Claude by adding a binance mcp, configuring a Python path, and using get price mcp to generate executive summaries emailed via Gmail.

Learn to debug MCPs with the inspector and MCP logs to diagnose integration issues, including production failures, by inspecting the Binance MCP log in VS Code.

Learn to integrate mic into Cursor by adding a global MCP server and configuring Binance MCP from a JSON config, then fetch crypto prices and email them with Cursor.

Debug MCP errors in cursor by inspecting the console behind cursor, using the developer tools, and reviewing extension host messages to locate MCP server issues.

Explore how resources extend LMS context with the MCP protocol, including listing and reading resources, templates for parameterized resources, and examples like crypto price data.

Implement activity logging for the MCP server by writing requests and errors to an activity log file with Pathlib, and expose logs as MCP resources via the file protocol.

Learn to use the MCP CLI from the Python MCP package to start a development server, run the MCP inspector, or install a server in the cloud desktop app.

Explore implementing and debugging resources and resource templates in the MCP bootcamp, connect to the MCP inspector, execute a crypto price workflow, and verify results via the activity log.

Explore resource management in cloud desktop by configuring the resource implementation path, adding a Binance MCP resource, restarting cloud, and reviewing the activity log for successful requests and errors.

Explore prompts as reusable, customizable instructions that shape how LLMs review code, tutor languages, and deliver executive summaries by packaging presets for tools and servers.

Design and use an executive summary prompt to fetch BTC and ETH prices and 24-hour percentage change, using get price and price change tools within the MCP workflow.

Explore how parametrized prompts with a crypto name parameter retrieve the current price and price change, using symbol mappings and prompt substitution in the MCP framework.

Learn to access MCP from Python by creating a stdio MCP client, initializing a session, listing tools, invoking a tool with arguments, and handling errors via the is_error flag.

Explore LangChain and LangGraph as open-source frameworks that simplify LM workflows with automatic tool calls. Learn to implement MCP access using LangChain’s tool routing and agentic behavior.

Learn how to integrate OpenAI with LangChain by loading the dot env values, configuring a deterministic chat model with temperature zero, and building an OpenAI-based workflow for a graph project.

Configure multi-server MCP clients and integrate a Landgraf React agent within LangChain to fetch current prices of Bitcoin and Ethereum, letting the agent select tools and providing debugging insights.

Use Lamb-smith with LangGraph and LangSmith to trace and debug MCP agents, capturing runs, latency, tokens, costs, built-in tool calls, and OpenAI function integration.