Agentic AI

Design AI agents that plan, use tools, remember and work together, safely.

  • Duration50 hoursTraining hours, including guided labs
  • Modules11Each module ends with hands-on practice
  • ProjectsLabs + capstoneBuilt, reviewed and demonstrated
  • LevelIntermediate to advanced

Course overview

50 hours, 11 modules, one capstone

A 50-hour course on AI agents that plan, call tools, keep memory and work together, and on running them safely in production. It starts where the Generative AI course ends: learners build a tool-using agent from scratch, then move to LangGraph, multi-agent systems and the Model Context Protocol (MCP). It finishes with guardrails, evaluation, tracing and deployment.

Learning objectives

  • Explain what makes a system agentic, and when not to use an agent.
  • Implement tool calling and the ReAct loop directly against model APIs.
  • Build stateful agents in LangGraph with checkpoints and human approval steps.
  • Design multi-agent systems and connect agents to real systems through MCP.
  • Evaluate, trace, secure and deploy agents within cost and safety limits.

Who it is for

Developers who have built LLM or RAG applications and want to automate multi-step work in support, operations, research, data or software teams.

Prerequisites
Python programming. The Generative AI course, or experience with prompting, model APIs and RAG.
Level
Intermediate to advanced
Status
Coming soon. Join the waitlist to hear first when registrations open.
Duration
50 hours of training, including guided labs
Join the waitlist

What you will learn

Skills covered in this course

Builds an agent from scratch before using any framework.

Agent fundamentals

Tool callingReAct loopPlanningMemory

LangGraph

Stateful graphsCheckpointsHuman-in-the-loop

Multi-agent systems

SupervisorHierarchical teamsCrewAIAutoGen

Production

MCPGuardrailsTracingEvaluationDeployment

Syllabus

Module-by-module curriculum

11 modules, 50 hours in total. Every module ends with hands-on practice.

01
Module 1

From LLM apps to agents

3 hrs
  1. 1.1What makes a system agentictools, loops, planning and memory · workflows vs agents · levels of autonomy
  2. 1.2Choosing the right designwhen an agent helps · when a simple workflow is better · a quick recap of the GenAI prerequisites
02
Module 2

Tool calling from scratch

5 hrs
  1. 2.1Toolsfunction and tool schemas · parallel tool calls · returning results to the model
  2. 2.2The agent loopthe ReAct pattern · stopping conditions · error recovery and retries
03
Module 3

Planning and reasoning

4 hrs
  1. 3.1Planning patternsplan-and-execute · task decomposition
  2. 3.2Self-improvementreflection and self-critique · using reasoning models
04
Module 4

Memory and state

4 hrs
  1. 4.1Memory typesshort-term vs long-term memory · conversation summarisation · vector and episodic memory
  2. 4.2Statestate schemas · persisting state between runs
05
Module 5

Building agents with LangGraph

7 hrs
  1. 5.1Graph basicsstate, nodes and edges · conditional routing · loops
  2. 5.2Reliability featurescheckpoints and persistence · human-in-the-loop approval · streaming progress
  3. 5.3Patternsrouters · tool-calling agents · subgraphs
06
Module 6

Multi-agent systems

6 hrs
  1. 6.1Architecturessupervisor · hierarchical teams · hand-offs between agents
  2. 6.2Frameworks comparedLangGraph · CrewAI · AutoGen
  3. 6.3Coordinationshared state · resolving conflicts · keeping teams small
07
Module 7

Connecting agents to the world

5 hrs
  1. 7.1Model Context Protocol (MCP)MCP servers and clients · exposing company data and actions
  2. 7.2Common toolsweb search and browsing · code execution sandboxes · SQL and REST API tools
08
Module 8

Safety and guardrails

3 hrs
  1. 8.1Threatsprompt injection through tools and documents · data leakage
  2. 8.2Controlsleast-privilege tool design · approval gates · sandboxing · budget and step limits
09
Module 9

Evaluation and observability

3 hrs
  1. 9.1Evaluating agentstask success metrics · trajectory evaluation · regression test sets
  2. 9.2Observabilitytracing with LangSmith or Langfuse · cost and latency monitoring
10
Module 10

Deploying agents

3 hrs
  1. 10.1Servingagents behind a FastAPI endpoint · streaming responses
  2. 10.2Running in productionbackground jobs and queues · scheduling · retries and timeouts
11
Module 11

Capstone project

7 hrs
  1. 11.1Design the agent systembusiness workflow · tools and permissions · evaluation plan
  2. 11.2Build, evaluate and deployagents and tools · guardrails and tracing · evaluation report · deployment
  3. 11.3Demo and reviewlive demonstration · review feedback

Total duration: 50 hours. Training hours across eleven modules, including guided labs and the capstone project.

Practical learning

Learn by building

  • A framework-free agent that uses a calculator, web search and a database tool.
  • A LangGraph support agent with memory and a human approval step before refunds.
  • A research team of planner, researcher and writer agents that produces a cited report.
  • An MCP server that exposes files or tickets to an agent, with guardrails and tracing.

Projects

  • Mini project · Helpdesk triage agentClassify incoming tickets, look up answers, draft replies and escalate when unsure.
  • Mini project · Research crewA multi-agent team that researches a topic and writes a cited brief.
  • Capstone · choose oneA lead-qualification agent, an IT operations assistant or a data-analysis agent that writes and runs code, evaluated and deployed.

Outcomes

What you can do after the course

Design agent systems

Choose the right pattern, from a single tool loop to a multi-agent team.

Build with LangGraph

Build stateful, resumable agents with human approval steps.

Connect real systems

Give agents safe access to data and actions through MCP and APIs.

Run agents safely

Evaluate, trace and limit agents in production.

Tool callingLangGraphMulti-agent designMCPGuardrailsAgent evaluationDeployment

Register

Join the waitlist for Agentic AI

Registrations for this course aren't open right now. Leave your details and we'll email you first, with the batch dates, fees and next steps.

  1. 1Join the waitlist
  2. 2We email you as soon as registrations open
  3. 3You receive the batch dates, fees and next steps
Course brochure (PDF)Overview, the full syllabus with hours, projects and outcomes.Open the PDF

Registered & approved

  • MSME registeredUdyam-registered enterprise, Government of India
  • AICTE partnerApproved partner of the All India Council for Technical Education

Join the waitlist

Program: Agentic AI

Preferred mode

We'll reply by email or phone with the next steps. By registering you agree to be contacted about this program.

Questions about Agentic AI?

Email us with your background and goals and we'll help you choose the right course.

All courses

We use cookies to personalise your experience and to analyse our website traffic.
By clicking “Accept cookies”, you consent to our use of cookies.

View privacy policy