MCP : Agent Skill, Claude Code, GenAI Model Context Protocol

Explore the evolution of large language models from prediction to external tools, highlighting the limitations of early models and the need for the model context protocol.

Explore client-server architecture, including the roles of clients, servers, and internet protocols, and understand how Rest APIs and standards enable scalable, reliable tool integrations for AI agents.

Discover the model context protocol (MCP) architecture, including client apps, protocol, and server, to enable standardized, two-way communication and seamless integration of external services.

Explore the MCP server components—tools, resources, and prompts—and learn how tools execute actions like retrieving data and updating databases, while resources provide read access and prompts stay user controlled.

Explore MCP transport types such as stdio, http with SSE, and streamable http, and understand how they enable scalable client-server communication.

Explore how MCP flows data across server, client, and host using json-rpc 2.0 over transport layers like stdio and http with sse, coordinating resources, tools, context, and prompts.

Explore an end-to-end agent system architecture for analyzing Excel files with Claude Desktop. Learn how a server hosts tools and prompts, files, and structured tool execution to summarize data trends.

Explore a beginner-friendly, hands-on walkthrough of the model context protocol drawn from official documentation, showing how its components come together in an end-to-end example.

Install and configure dependencies for the MCP server using the UV package, create a venv, and sync dependencies to prepare the MCP tool for LM invocation.

Create an MCP server for weather forecast by integrating weather.gov endpoints, parsing forecast data for a given latitude and longitude, and using a user agent header.

Develop an MCP server to fetch weather data via a generic URL fetch with json response, user agent headers, timeout, and error handling, showcasing cursor IDE capabilities for generative AI.

Create an MCP tool to fetch weather forecasts using latitude and longitude, decorated with the MCP decorator, and handle async calls to grid points and forecast endpoints.

Create and register an MCP server for weather data, configure Claude for desktop, and run the get forecast tool to retrieve a five-day outlook.

Use the MCP inspector to test server configurations and endpoints before exposing to clients. Install node, run npx model-context-protocol/inspector to connect and verify the weather server v1.101 and proxy tokens.

Integrate a Claude desktop MCP server with GitHub by creating a GitHub token, configuring MCP server settings, and executing repository searches, pull requests, and merges on the cloud agent.

Use docker to host MCP servers from the GitHub catalog, accessing 134 available servers via the MCP toolkit with token-based setup and streamlined cloud desktop deployment.

Explore how to run the MCP server with Docker, interact with GitHub to create branches and pull requests, update a readme, and merge changes using Claude.

Implement an MCP protocol using SSE for real-time updates, with FastAPI and Starlette under an ASGI server, exposing streaming and JSON RPC endpoints via SSH transport, run on port 8100.

Create a weather client for the MCP server using SSC transport, then establish an SSH connection and initialize a client session to reveal the server name and version.

Connect to the weather server via ssh transport, list available tools, and call the forecast method exposed as an MCP tool with latitude and longitude to retrieve tonight's temperature.

Expose the MCP server over streamable http by importing transport types, parsing arguments, and running uvicorn on port 8123, while the streamable http app manages session, routing, and authentication.

Demonstrates invoking the MCP server with the inspector over streamable http, connecting, listing resources, prompts, forecasts, and alerts, then using a Python client to fetch endpoint results.

Learn to build an MCP client for a streamable HTTP server in Python, establishing async connections, managing sessions, and listing server tools like get forecast and get alerts.

Discover how prompts structure instructions for generative AI, using instruction, context, input data, and output format to translate human intent into AI action.

Explore zero-shot, few-shot, and chain-of-thought prompting in amazon bedrock through the prompt management console, testing prompts with variables, and configuring models like claude 2.1 for consistent results.

Expose well crafted prompts through the MCP server prompts, using a weather advisor prompt decorator and an async method to generate location-based weather advice for activities and planning.

Build an MCP client in Python to connect to a streamable HTTP server, initialize the client session, and list tools such as get forecast and get alerts.

Create a resource endpoint by locating application logs, decorating the resource, and returning a string with header and date-time context, then start the server with stdio transport.

Update the MCP resource configuration to use the new weather resource file, load app logs, and demonstrate how Lem analyzes critical issues like high CPU, timeout errors, and database performance.

Observe how MCP resources update data on the fly as logs grow, with a weather app example showing new lines appearing after refreshing. Reopen to trigger updated resource data.

Explore the MCP inspector to list exposed resources and inspect logs, endpoints, and URIs. Use the inspector to connect, test endpoints, and verify your application works before exposing resources.

Explore cloud code, an AI assistant by Anthropic optimized for programming tasks that enables collaborative code writing, debugging, and refactoring across the development lifecycle.

Install cloud code using npm, set up a project with cloud code and subagents, and configure login, API token, and CLI settings to enable cloud usage.

Discover how cloud code can analyze and navigate code, run task specialized agents, automate documentation, and support multiple languages, while learning to use slash commands to fetch and summarize documentation.

Learn to scaffold a Python hello world project with Claude Code, creating main.py, readme, and requirements.txt, then run python3 main.py to print hello world and explore cloud code workflows.