Learning Series / How to Build Agents

How to Build Agents

A hands-on, six-part walk from a blank folder to a deployed, specialized agent — setup, architecture decisions, tools and memory, deployment, and the multi-agent scaling most courses skip entirely. Based on the written curriculum at shreekantmandvikar.com/writing/agentic-ai-curriculum.

AgentsGenAIHands-On

6 parts, top to bottom

  1. Setup Guide

    Setup Guide

    Python and the OpenAI API, set up in three steps — installing Python, the required libraries, and your first API call.

    Read the full post →

  2. A Practical Roadmap

    LangGraph or crew.ai? Single agent or multi-agent? The real decisions to make before writing a line of code, including where MCP fits in.

  3. Add Tools & Memory

    Add Tools & Memory

    To make an AI agent truly intelligent, it needs two abilities: tools to act, and memory to think. Tools let it perform real actions — fetching live data, sending a message. Memory lets it recall past turns, preferences, and context, turning a static chat into something that actually keeps up.

    Read the full post →

  4. Deploy Your Agent

    Deploy Your Agent

    Your prototype is ready — now what? Building a brilliant agent is the easy part; deploying it into a live environment is where the real value (and the real challenge) actually is. A structured path from proof-of-concept to something you can trust in production.

    Read the full post →

  5. Specialized Agents

    Specialized Agents

    This is where agents stop being generic assistants and start becoming powerful, outcome-driven systems. Specialized agents are designed for a single purpose and deliver clarity, speed, and consistent execution.

    Read the full post →

  6. Scale and Automate

    From one agent to a system of them — deployment, observability, and what changes once more than one agent is running.

← back to Learning Series