Agentic AI: Build AI Agents with LangGraph, CrewAI & MCP

Install and set up the development environment by installing Visual Studio Code and the UV package manager, with guidance for Windows, macOS, and Linux, to manage Python packages.

Introduce uv, a Rust-based Python package manager that auto manages virtual environments, tracks dependencies with log files and a TOML project file, and supports init, add, lock, sync, and run.

Discover how to build your first agentic AI workflow in a Jupyter notebook using Python and Deep Sea or OpenAI APIs, with environment variables and a semi-agentic loop.

Explore how ai agents use llm outputs to control workflows, perceive environments, and autonomously decide and act toward goals, highlighting the contrast with traditional predefined workflows.

Discover LLM workflow design patterns to turn large language models into reliable systems, including prompt chaining, routing patterns, and task decomposition with context management.

Learn the parallelization design pattern, coordinating independent LM tasks with a coordinator, subtasks, and an aggregator to reduce latency and boost performance.

Understand the agent design pattern as an autonomous, goal-directed ai system. Learn how goals, planners, reasoners, tool sets, and optional memory enable reasoning, action, observation, and iterative goal achievement.

Compare top LLMs for real-world autonomous agents, including paid and open-source options, and their fit for local versus cloud execution and planning, memory, and tool use.

Explore how resources and tools boost LLM capabilities in agentic AI, using data context, retrieval augmented generation, and tool chaining to execution engines.

Build a chatbot that acts like you using a Jupyter notebook and Gradio interface, leveraging a system prompt and your professional and personal summaries to respond on your behalf.

Learn how Pydantic enforces type safety and data validation, uses Python type annotations, and automatically converts inputs while handling settings management for robust LLM workflows.

Build a multi-line pipeline to evaluate an LLM output via a second LM using the evaluator and optimizer pattern. Implement in a notebook with pedantic schemas and Gradio live chat.

In a multi-llm pipeline, build a chatbot that reevaluates deep seq outputs with Gemini 2.0 via an evaluator-optimizer loop until acceptable.

See how intelligent agents perform function execution and respond to alerts using LLMs, integrating push notifications and tools for real-time automated trading scenarios.

Learn how LLMs handle tool calls and function calls by defining tool schemas, dispatching with a tool dispatcher, and routing user details and unknown questions to mobile notifications.

Expand ai assistants to handle unknowns with tooling by replacing if-else with a dynamic tool dispatcher using a function dictionary. Explore testing, globals vs. tool maps, and system prompts.

Build an end-to-end ai agent that uses a dispatcher to route tool calls to predefined tools, enabling interactive conversations, tool use, and mobile notifications.

Learn to build an autonomous AI assistant using DeepSeek with the OpenAI SDK to handle unknowns across different language models, capture user details, and notify the owner for deployment.

Deploy your digital avatar ai agent as a Gradio app from app.py, loading environments and tools, then publish via Docker, Kubernetes, or Hugging Face Spaces for a public URL.

Explore Python async IO, a single-threaded, non-blocking event loop for I/O bound tasks, and compare with multithreading and multiprocessing within the OpenAI SDK powered LMS workflows.

Explore the OpenAI agent sdk, its core primitives—agents, handoffs, guardrails—and how to build modular, hierarchical agent systems with tool calls, input validation, and end-to-end tracing.

Explore the three building blocks of the OpenAI agent SDK—agent, runner, and trace—and learn how these components handle prompts, memory, tools, and execution with any llm.

Create multiple AI powered sales outreach agents with the OpenAI agent SDK, each with distinct tones (professional, witty, concise) and streaming emails for automated business outreach.

Learn to generate multiple emails in parallel with professional, engaging, and concise agents using async io, then evaluate and select the best email with an ai selector and trace.

Orchestrate tools with agents for ai sales execution, wrapping functions and agents as tools, set up sendgrid, and empower a sales manager to coordinate three agents and send the email.

Differentiate agent as a tool from handoffs in multi-agent architectures, and build a hybrid workflow with a subject-line generator, html formatter, and an email manager.

Examine a hybrid multi-agent workflow using tools and handoffs to generate and format HTML sales emails, coordinated by a sales manager and finalized by an email manager.

Explore multi-model integration with Gemini Deep Sea Grok and OpenAI agents, routing outputs, performing agentic evaluation, tool-based handoffs, and guardrails to generate professional HTML emails.

Master multi-model integration with Gemini, DeepSeek, and Groq to orchestrate OpenAI agents that generate email subjects, format content in HTML, and send messages via SendGrid.

Demonstrate multi-model integration across Gemini, DeepSeek, and Grok with OpenAI agents, executing a cross LM sales workflow through tool orchestration, email formatting, and traceable evaluation.

Implement guardrails and structured output to validate and secure AI agent systems. Use Pydantic schemas for JSON outputs to ensure regulatory compliance, auditable and deterministic results.

Build a supercharged web agent with real-time web search to mine the internet and return a clean, structured summary for business research, competitive analysis, and real-time intelligence.

Plan web searches with precision using a pedantic Pydantic schema to generate a structured, reasoned set of queries for a user query.

Design an end-to-end asynchronous research pipeline that executes planner-generated queries in parallel, creates a markdown report, and emails the results via a dedicated agent flow.

Design and execute an end-to-end autonomous research pipeline that performs parallel web searches using async IO, generates a structured report with a writer agent, and dispatches it via email.

Transform end-to-end research notebooks into a modular, production-ready AI system with plug-and-play components. Leverage Gradio to deploy an interactive interface for planner, web search, writer, and email agents.

Create a gradio powered app.py with a customized UI using blocks to run the financial research agent pipeline, loading .env variables and displaying the final markdown report.

Explore how Crew AI orchestrates multi-agent teams with defined roles, tools, and memory, and compare its collaboration model to the OpenAI SDK, including open-source and enterprise offerings.

Demystify crew ai by unpacking agents, tasks, and the crew, and explore how sequential, concurrent, and hierarchical processing modes orchestrate their interaction.

Learn to integrate multiple llms into a crew ai project using light lm, with centralized rate limits, logging, and failover, plus project setup and yaml configuration for agents and tasks.

Design and implement a think bot within a fresh crew I project, building an argue bot and verdict bot to handle make case, break case, and render verdict tasks.

Learn to configure and orchestrate agents and tasks with crew.py and main.py, loading configurations from agents YAML and task YAML to drive think bot and verdict bot debates.

Learn to build a financial agent with crew ai by adding tools and context, leveraging a google search tool via serp api (sirpur) for real-time data.

Configure two agents, the financial research expert and market insight analyst, to build a crew ai research tool that produces an investor-ready market report via sequential tasks with context passing.

Learn to empower autonomous agents with real-time web search by integrating the surfer dev tool and server queries, ensuring up-to-date data beyond model training.

Build a multi-agent ai-powered investment picker called stock whisper, evaluating stock fundamentals and technicals, using structured pedantic json outputs and a hierarchical workflow to deliver actionable recommendations.

Set up the Profit Pilot project in Visual Studio Code, configure market trend, deep dive, and equity selector agents, and define tasks to identify market movers and analyze fundamentals.

Define structured output with pedantic schemas for company data, and orchestrate a multi-agent crew ai workflow including market trend tracker, deep dive researcher, and equity selector with a delegating manager.

Explore toolsmith mode to craft custom alerts and json tools for Crew AI. Build a push notification tool and use structured outputs to present top investment picks with confidence scores.

Equip crew AI with memory using vector search and SQL recall to enable memory-based agents, covering short term to long term, contextual, entity, and user memories.

Equip crew ai with long-term and short-term memories, including entity memory, using vector search and sql recall; implement embeddings, rag storage, and memory retrieval to enhance agent performance.