Learn how AI agents perceive, reason, plan, and act
A free, hands-on introduction to Agentic AI. Explore interactive diagrams, run Python examples right in your browser, and finish with a clear picture of how to build agents that are useful and safe.
What you will learn
This crash course is offered by the Agentic AI @ KFUPM community. It assumes no prior knowledge of Agentic AI.
Learning objectives
- Explain how AI evolved from GANs and Transformers to agents and Physical AI.
- Describe the components of an agent and how they work together.
- Choose a suitable design pattern for a given task.
- Explain how always-on agents work in the background and stay under human control.
- Connect agents to tools and data with MCP and decision APIs.
- Build and trace an end-to-end agent loop in Python.
- Prototype workflows with no-code platforms such as n8n.
- Build agents in Python with SDKs and reach many models through OpenRouter.
- Apply guardrails, evaluation, monitoring, and human oversight.
Who it is for
University students, researchers, faculty, and practitioners from any discipline who want a practical understanding of AI agents.
Prerequisites
- Basic Python: variables, functions, lists, and dictionaries.
- A web browser. Most examples run in the page; others open in free Google Colab.
- No paid API keys are required. Module 9 uses a free OpenRouter key.
How to use this site
Work through the modules in order or jump to what you need. Each module has an interactive diagram, short explanations, runnable code, and a quick check of your understanding.
Ten modules, from foundations to governance
Modules 1 and 2 build the vocabulary. Modules 3 to 9 show how to design, connect, and run agents, from patterns and always-on agents to no-code and code-based workflows. Module 10 covers how to deploy them responsibly.
Evolution of AI
From GANs and Transformers to multimodal models, agents, and Physical AI.
Components of AI Agents
Perception, reasoning, planning, action, tools, memory, and oversight.
Agent Design Patterns
Sequential, routing, parallel, tool calling, reflection, planning, and multi-agent.
Always-on AI Agents
Proactive agents that work in the background: dots, Grok Bot, Instinct, and Muse.
Model Context Protocol
A standard way to connect agents to tools, data, and prompts.
Decision APIs
Rules engines, optimizers, and OpenAPI decision services as agent tools.
Agentic AI Workflows
Perception to feedback: an end-to-end agent loop you can run.
No-Code Agentic Workflows
Build agent workflows visually with n8n and similar platforms.
Code-based Agentic Workflows
Agent SDKs in Python, and many models through OpenRouter and Commonstack.
Agentic AI Governance
Guardrails, evaluation, monitoring, and human oversight.
Labs
Every Python example in the course, grouped by module. Examples marked Runs in browser run directly on the module page with one click. The others open in Google Colab because they install packages.
Enable JavaScript to see the list, or open any module page. All notebooks are in the notebooks folder of the repository.
By the end you will be able to...
- Read an agent architecture diagram and name each part.
- Sketch a workflow for a real task and justify the pattern you chose.
- Write a small agent loop in Python with tools, memory, and feedback.
- Expose a capability to agents through an MCP server or a decision API.
- Choose an agent SDK and switch between free and commercial models.
- Propose guardrails, an evaluation plan, and an approval policy for an agent.