MCP Mastery: Build AI Apps with Claude, LangChain and Ollama

Explore loop engineering and how a react agent uses an action–observe cycle, tool calls, and a two-model, self-prompting setup with external memory sharing to improve outputs.

Explore the MCP architecture with host, client, and server components and JSON-RPC 2.0 data flow, plus its versioned evolution and features like resources, prompts, and tooling.

Explore the MCP architecture, where a host and client connect to a server to access tools, prompts, and resources, reducing cross-service connections and illustrating linear vs square complexity.

Explore the fast MCP overview, detailing host-client-server architecture, stdio and streamable http transport layers, and how a server integrates local, database, and internet operations.

Explore the MCP transport layer, comparing stdio for local servers with streamable http for remote, real-time bidirectional streaming, and wire tools, resources, and prompts through the MCP server.

Build a first MCP server and client demo in Python using server dot pi and client dot pi. Show how the host, client, and server connect to initialize a session.

Build an MCP client with the default MCP Python SDK, connect to the server via streamable http, and compare it with the fast MCP and a Python SDK-based client.

Build your first MCP client with the FastMCP SDK, connect to a server, list available tools, and call tools to perform operations like adding two numbers.

Build a real-time weather MCP server and client to fetch current conditions and a three-day forecast for a city, using an open web source and an async Python client.

Explore ready-made MCP servers and test your own MCP server with the MCP inspector. Connect via localhost, run tools, and explore resources, prompts, and official integrations.

Use the MCP inspector with the stdio transport to run the MCP server, configure paths, and test tools like get weather and get forecast.

Connect your MCP server to a local agent using Ollama to host and serve the model, install Ollama, pull a model, and run the MCP usage workflow locally.

Learn to connect an LM to an MCP server with LangChain and Ollama, install line chain and MCP libraries, and build a weather server with an agent client.

Configure the MCP server and agent client by setting the MCP server name, the run command, and the working directory, then address transport issues for proper tool calls.

Pull the Quinn three 8 billion model from Ollama and connect it to the agent client. Execute the agent to fetch current weather in New York and a three-day forecast.

configure the advanced mcp agent client to read mcp config from a json file and run multiple servers via mcp.json. reference config path and invoke addition, multiplication, subtraction, and division.

Configure a two-server MCP setup by adding a math server and a weather server in the servers directory, update MCP JSON, enable server manager, and run the advanced agent client.

Execute an MCP agentic client with multiple servers by identifying MCP server tools, using GPT OS (20 billion) to call them, including get forecast and get wither for New York.

Configure MCP servers with streamable-http transport and stdio, set mcp http json for weather and match services on ports 8000 and 8001, and run the advanced agent client.

Learn to run an MCP server on cloud desktop by pasting self-written math and weather servers into cloud desktop, enabling studio mode, and configuring ports 8000 and 8001.

Learn to run local MCP servers on Claude Desktop, use weather and math tools (get weather, get forecast, add, multiply, subtract, divide), and manage tool permissions for seamless AI queries.

Learn to automate web browsers with the Microsoft Playwright MCP server on the cloud desktop, install using npm, and run tasks like navigation, search, and page snapshots.

Set up and use the Airbnb MCP server to search hotel listings near Powai, Mumbai, and retrieve listing details via the MCP tools.

Set up MCP servers for data analysis and use the Excel MCP server to read and write Excel data, connect to SQL and the file system, via cloud desktop.

Learn to analyze a support ticket Excel sheet with the MCP server on a cloud desktop, implementing color coding and generating a data analysis sheet.

Configure and run the MCP server to analyze Excel data and automatically generate a detailed, colorful PowerPoint presentation, including local and cloud installation steps.

Create a comprehensive PowerPoint from support ticket data using the PPT and Excel MCP servers, populating slides with charts, tables, bullet points, and an action plan.

Learn to create a pdf report by connecting Notion with the MCP server, set up the Notion API key and integration, and configure access in the cloud desktop.

Set up a Notion API integration by creating a new integration, retrieving and pasting the secret key, and granting read, write, and insert update access to connect pages and reports.

Connect to the Notion server, upload the Excel data, and have the AI generate a colorful support ticket data analysis report with executive summary, metrics, and a PDF export.

Set up the MCP server with SQLite and attach it to cloud desktop to access the SQL database, then download, extract sample database, and verify the SQL server is running.

Connect cloud desktop to the escalate MCP server and perform SQLite database analysis with AI, running queries to find top salaries, department averages, and manager counts.

Explore how to connect and configure the file system MCP server, manage read and write operations, and organize downloads by granting access to specific directories via cloud desktop.

Explore how the MCP framework uses tools, resources, and prompts to power a personalized job search from a resume on cloud desktop to final results.

Explore MCP resources and types, including static and dynamic resources, text and binary data, URIs, mime types, and metadata, and learn duplicate resource management with prompts.

Explore how MCP prompts act as a user-controlled guidance layer. Learn about centralized server-side prompt repositories, dynamic resources and tools, and simple to structured prompt patterns.

Differentiate tools for real-time operations, resources for static context, and prompts for guided workflows; illustrated by a job search assistant using all three layers.

Set up the MCP server with server.py, create directories for jobs, temp, and resume storage, and load API keys from a dot env file to enable job search integration.

Learn to build a job search API with Rapid API, obtain an API key and free plan access, and test endpoints using curl and environment variables.

Implement the search jobs tool in the MCP server using Rapid API, handle credentials, fetch up to five job results, summarize descriptions, and store them in a temporary directory.

Implement the save_job tool to move a selected job from the temp data into the candidate saved folder, extracting or inferring salary and storing a structured json.

Create two MCP resources: a resume resource that reads the resume.pdf with PyPDF2 and returns markdown, and a saved job resource that compiles saved job JSONs into a markdown listing.

Develop MCP prompts to analyze the job market, generate personalized job recommendations, and create match reports that compare resume details with saved job data.

Test and run the job search MCP server on cloud desktop, configure the environment, and use the job search tool to analyze the job market for a role and location.

Explore testing a full fledged job search AI app using MCP prompts, resources, and tools to save jobs, run resume matching, and generate personalized job recommendations with a match report.

Build a rag system with MCP by ingesting PDFs into a chroma DB and creating embeddings with lang chain and lama. Explore gateway architecture and search across cloud and desktop.

Ingest PDFs from files, folders, or URLs and extract text with PyPDF2. Split text into chunks using LangChain, generate embeddings, and store them in a Chroma DB for similarity search.

Configure the MCP RAG server by writing server.py, installing dependencies via a virtual environment from pyproject.toml, and wiring LangChain with Ollama embedding for a vector store, plus PDF text extraction.

Implement PDF text extraction from a PDF file, folder, or URL using PyPDF2, split text into chunks with a recursive character splitter, and ingest embeddings into a chroma vector store.

Download a pdf from a url into a download directory, extract text, and ingest it using MCP tools for folder ingestion and retrieve db info and clear db.

Implement the ingest pdf mcp tool to handle pdf files, folders, and remote urls, then extract, create embeddings, chunk, and store vectors with error reporting.

Implement retrieve MCP tool to query chroma vector store, run a similarity search for top chunks with scores and metadata, and expose db_info and clear_db tools for database management.

Set up and run the MCP rag tool with chroma db, embedding models, and server.py while verifying llama status, then use tools to ingest pdfs and manage the db.

Ingest data with MCP rag on a cloud desktop, index PDFs into chroma db rag using 4096-byte chunks, then verify DB info and retrieval across multiple sources.

Build a Streamlit app to chat with your MCP rag using LangChain MCP adapters, set up a client-server connection, load MCP tools, and configure an agent.

Build a Streamlit app for an MCP client using the llama and the queen three model, with async response generation, tool integration, and a debug view to trace chat history.

Trigger the submit button to generate a response, store and display chat history in reverse order, and run MCP RAG with LangChain and Ollama via a Streamlit app.

Explore building a research assistant with MCP and LangGraph, wiring a multi-server MCP, line graph agent, Firecloud web scraper, memory, and chroma vector store for semantic search.

Initialize a fast MCP server, configure chroma DB for storing research topics, and set up embeddings with llama and Ollama for document ingestion.

Learn to generate and store hash values for content to detect duplicates, save hashes in content hash JSON by topic, and manage Chroma DB embeddings for topic-specific vector stores.

Implement the save_research_data MCP tool to ingest and deduplicate content, store it in a Chroma DB vector store under the topic path, and track content hashes for future saves.

Implement the search research data MCP tool to retrieve topic-specific vectors from a chroma db, perform a similarity search, format results, and handle errors.

Implement and test MCP tools to list, delete, and get info on research topics using chroma db, topic directories, and vector store metadata.

Test your MCP server with Claude using the cloud desktop, verify the research server and MCP tools, and confirm topics and stock research data store in the topic db.

Learn to configure API keys for FireCrawl and LangSmith, set up the MCP server and environment variables, and wire client.py with the LangChain MCP adapter and Ollama for AI apps.

Import libraries and asyncio, load environment variables including the fire API key, configure a multi-server MCP client with a JSON file, and prepare the research agent for graph-based state management.

Build a create research agent using LangGraph, binding tools to an LLM, and configuring a line-graph driven chat and tool node workflow with conditional edges and memory for knowledge retrieval.

Build a LangGraph MCP client with a command line interface that dynamically retrieves tools from the MCP server and handles async agent responses.

Test and demonstrate a research assistant connected to MCP using LangGraph, LangSmith logging, and in-memory storage, showing topic management, tool calls, and saving results to a vector DB.