AgenticAI Foundry 🤖

MIT Professional Education: Applied Generative AI for Digital Transformation Interactive demos for understanding AI economics, multi-agent systems, and agent integration

---

📥 New Here? Start With This

⬇️ Download the Student Quick Start Guide (PDF)

The Quick Start Guide walks you through everything with screenshots — how to get the code, which setup path to choose, and what you should see when it's working. Download this before doing anything else.

---

🎯 What's Included

Demo	Module	What You'll Learn	API Key Needed?
💰 LLM Cost Explorer	Module 1	Why the same AI task can cost $1 or $230 depending on model choice	No
🤖 Multi-Agent Demo	Module 2	How three AI agents collaborate like a team (CrewAI)	Optional
🔗 LangChain Agent Demo	Module 2	How a single agent uses tools to answer questions in real time	Optional
🔌 MCP Explorer	Module 3	How AI agents connect to external tools (calendars, CRMs, databases)	No
🛡️ Agent Security Demo	Module 4	Prompt injection attacks and defense-in-depth guardrails	Demo: No / Live: Optional

Modules 1, 3, and 4 (Demo Mode) work immediately — no account, no API key required. Module 2 demos need either Ollama (free, local) or an OpenAI API key.

---

🚀 Getting Started — Choose Your Path

Not sure which path to take? Ask yourself one question:

"Have I used Docker before, or am I comfortable installing a new application from a website?"

• Yes → Take Path A: Docker — more reliable, fewer things can go wrong once running
• No → Take Path B: Python — more familiar territory if you've used Python before

Both paths give you the full app. Docker is recommended because it handles all the dependencies for you.

---

Step 0 — Get the Code (Both Paths)

You need a copy of this repository on your computer before doing anything else.

Option 1: Download ZIP (easiest — no GitHub account needed)

1. Click the green < > Code button at the top of this page
2. Click "Download ZIP"
3. Find the downloaded file (usually in your Downloads folder)
4. Right-click → Extract All (Windows) or double-click (Mac) to unzip it
5. Remember where you saved the folder — you'll need it in the next steps

Option 2: Clone with Git (if you have Git installed)

git clone https://github.com/dlwhyte/AgenticAI_foundry.git

---

Path A: Docker (Recommended)

Docker packages the entire app — Python, all libraries, everything — into a self-contained box. Once installed, it runs identically on Windows, Mac, and Linux with no dependency conflicts.

What is Docker? Think of it like a shipping container for software. Everything the app needs is packed inside. You don't need to install Python or any libraries separately.

Time required: ~20 minutes first time (mostly waiting for downloads). Under 1 minute after that.

Step 1 — Install Docker Desktop

• Windows or Mac: Download from docker.com/products/docker-desktop and run the installer
• Linux: See docs/DOCKER_GUIDE.md for Linux-specific instructions

After installing, open Docker Desktop and wait for it to fully start (you'll see a whale icon in your taskbar/menu bar stop animating).

✅ You'll know Docker is ready when the whale icon is still (not animating).

Step 2 — Open a Terminal

System	How to open a terminal
Windows	Press Win + R, type powershell, press Enter
Mac	Press Cmd + Space, type Terminal, press Enter
Linux	Press Ctrl + Alt + T

Step 3 — Navigate to the Project Folder

Type the following, replacing the path with where you saved the folder in Step 0:

# Windows example (adjust the path to match where you saved it)
cd C:\Users\YourName\Downloads\AgenticAI_foundry

# Mac/Linux example
cd ~/Downloads/AgenticAI_foundry

💡 Tip: You can drag the folder onto the terminal window and it will paste the path for you.

Step 4 — Build the App (one-time only)

docker build -t agenticai-foundry .

This downloads Python and all required libraries. It will take 2–5 minutes and show a lot of output — that's normal. Wait until you see:

Successfully built xxxxxxxxxx
Successfully tagged agenticai-foundry:latest

Step 5 — Run the App

docker run -p 8501:8501 agenticai-foundry

You'll see output ending with something like:

You can now view your Streamlit app in your browser.
URL: http://localhost:8501

Step 6 — Open in Your Browser

Open your web browser and go to:

http://localhost:8501

🎉 You should see the AgenticAI Foundry home page!

Step 7 — Stop the App

When you're done, go back to the terminal and press Ctrl + C.

Troubleshooting Docker

Problem	Solution
"Docker command not found"	Make sure Docker Desktop is open and running
"Cannot connect to Docker daemon"	Open Docker Desktop and wait for the whale to stop animating
"Port 8501 is already in use"	Run docker run -p 8502:8501 agenticai-foundry and open http://localhost:8502
Build seems stuck	Wait — first build downloads ~500MB, can take 5+ minutes on slower connections

For more detailed Docker troubleshooting, see docs/DOCKER_GUIDE.md.

---

Path B: Python

Run the app directly using Python. You'll have more visibility into the code but there are more steps that can vary by operating system.

Time required: ~15 minutes, more steps involved.

⚠️ Check your Python version before starting. Open a terminal and run:

python3 --version

If it shows Python 3.9 or lower — stop and use Docker (Path A) instead. CrewAI (required for Module 2) only works on Python 3.10 or higher. Modules 1 and 3 will still work on Python 3.9, but Module 2 will not.

Step 1 — Install Python 3.10 or Higher

Check your version first:

python3 --version

• If you see Python 3.10, 3.11, 3.12, or 3.13 — skip to Step 2, you're ready
• If you see Python 3.9 or lower — download a newer version from python.org/downloads and run the installer

⚠️ Windows users: During installation, check the box that says "Add Python to PATH" — this is easy to miss and causes problems later if skipped.

⚠️ Mac users: Your Mac may have Python 3.9 built in (from Xcode). If so, download Python 3.11 from python.org/downloads — it installs alongside the existing version without breaking anything.

Step 2 — Open a Terminal and Navigate to the Project Folder

See Step 2 and Step 3 from the Docker path above for how to open a terminal and navigate to the folder.

Step 3 — Install the App's Dependencies

⚠️ This step is required — do not skip it. If you skip this, the app will crash with an error like ModuleNotFoundError: No module named 'tiktoken'.

Run both of these commands, one at a time:

pip3 install -r requirements.txt
pip3 install -r requirements-crewai.txt

This downloads all required Python libraries. It will take 2–5 minutes and show a lot of output — that's normal. Wait until you see your terminal prompt again before moving on.

💡 If pip3 gives "command not found", try pip install -r requirements.txt instead.

Step 4 — Run the App

python3 -m streamlit run Home.py

Your browser should open automatically to http://localhost:8501. If it doesn't, open your browser and go there manually.

Step 5 — Stop the App

Press Ctrl + C in the terminal.

Troubleshooting Python

Problem	Solution
"pip not found"	Try pip3 install -r requirements.txt instead
"streamlit not found"	Try python -m streamlit run Home.py
"Permission denied"	Add --user to the pip command: pip install --user -r requirements.txt
Browser doesn't open	Manually go to http://localhost:8501 in your browser

---

🤖 Setting Up the Agent Demos (Module 2)

The LLM Cost Explorer (Module 1) and MCP Explorer (Module 3) work without any additional setup.

For the Multi-Agent Demo and LangChain Agent Demo, the app needs an AI model to run. You have two options:

Option A: Ollama — Free, Runs on Your Computer

Ollama lets you run AI models locally. No account, no cost, no data leaving your machine.

Step	Command
1. Install Ollama	Download from ollama.ai
2. Download a model	ollama pull llama3.2 (2GB download, one-time)
3. Start Ollama	ollama serve — keep this terminal open

Then run the app (Docker or Python path) and select Ollama in the demo settings.

💡 The model download is large (~2GB). Do this on a good internet connection.

Option B: OpenAI — Paid, Faster Results

Costs approximately $0.01 per demo run. Requires creating an account at platform.openai.com.

Once you have an API key, enter it directly in the app's sidebar — no environment setup needed.

For full setup instructions for both options, see docs/CREWAI_SETUP.md.

---

📚 Documentation

Guide	What It Covers
📥 Student Quick Start (PDF)	Screenshots walkthrough — start here if you're new
🐳 Docker Guide	Full Docker setup with troubleshooting
📖 Beginner's Guide	Deep explanation of all technologies used
💰 LLM Cost Guide	Token economics and model pricing (Module 1)
🤖 Multi-Agent Guide	CrewAI vs LangChain patterns (Module 2)
🔌 MCP Guide	Model Context Protocol explained (Module 3)
⚙️ CrewAI Setup	Ollama and OpenAI setup for agent demos

---

📁 Project Structure

AgenticAI_foundry/
├── Home.py                          # Start here — the course hub
├── pages/
│   ├── 1_LLM_Cost_Calculator.py    # Module 1: Cost explorer
│   ├── 2_Multi_Agent_Demo.py       # Module 2: CrewAI multi-agent
│   ├── 3_LangChain_Agent_Demo.py   # Module 2: LangChain single agent
│   ├── 4_MCP_Explorer.py           # Module 3: MCP protocol
│   └── 5_Agent_Security_Demo.py    # Module 4: Prompt injection & guardrails
├── crews/                           # Multi-agent logic (CrewAI)
├── agents/                          # Single-agent logic (LangChain)
├── docs/                            # All guides and documentation
│   ├── Student_Quick_Start.pdf      # ← Start here if you're new
│   ├── DOCKER_GUIDE.md
│   ├── BEGINNERS_GUIDE.md
│   └── ...
├── setup_check.py                   # Run this to check your environment
├── Dockerfile
├── requirements.txt
└── README.md

---

🔍 Check Your Environment

Not sure if everything is set up correctly? Run this from your terminal in the project folder:

python setup_check.py

It will check Python, required libraries, Docker, Ollama, and API keys — and tell you in plain English what's working and what needs attention.

---

🛠 Technologies Used

Technology	What It Does
Streamlit	Creates the web interface
CrewAI	Orchestrates multiple AI agents
LangChain	Connects to AI providers and tools
Ollama	Runs AI models locally on your machine
Docker	Packages the app for reliable deployment
Plotly	Interactive charts and visualizations

---

❓ Getting Help

• Setup issues: Run python setup_check.py and share the output
• Docker problems: See docs/DOCKER_GUIDE.md troubleshooting section
• Agent demo issues: See docs/CREWAI_SETUP.md troubleshooting section
• General questions: Post in the course discussion forum with any error messages you see

---

📄 License

MIT License — see LICENSE

---

MIT Professional Education | Applied Generative AI for Digital Transformation Modules 1, 3 & 5 (Demo Mode) work immediately with no API key · Module 2 requires Ollama or OpenAI