1. What This Means in the Exam
Using Azure AI Foundry, you:
Create an agent in the portal
Then build a lightweight client app that:
Sends user input
Receives responses
Displays results
👉 Think:
“A simple front-end for your AI agent”
🤖 2. What is a Lightweight Client Application?
A lightweight client is typically:
A console app
A simple web page
A basic chatbot UI
It does not contain the AI logic itself—it just calls the agent.
🔄 3. How It Works (Core Flow)
User → Client App → Agent (in Foundry) → Model/tools → Response → Client App → User
🧩 4. Key Steps (Exam-Friendly)
1️⃣ Create the Agent
In Azure AI Foundry:
Choose model (e.g. LLM)
Define instructions
Add tools or data if needed
2️⃣ Deploy the Agent
Makes it accessible via an endpoint/API
3️⃣ Build the Client App
Using an SDK (e.g. Python, C#, REST):
The app:
Connects to the endpoint
Sends a message (prompt)
Receives a response
4️⃣ Display the Response
Show output in UI (console/web/chat)
💬 5. Example (Conceptual)
User input:
“Book a meeting tomorrow at 10am”
Flow:
Client sends request
Agent interprets request
Agent may call a tool (calendar API)
Response returned:
“Meeting scheduled for tomorrow at 10am”
🎯 6. When This is Used
Chatbots in websites
Internal assistants
Automation tools
AI copilots
⚖️ 7. Key Exam Distinctions
Feature Client App Agent
Role Interface Intelligence
Contains AI logic? ❌ No ✅ Yes
Calls model/tools? ❌ No (indirectly) ✅ Yes
Built with SDK? ✅ Yes Created in portal
⚠️ 8. Common Exam Traps
❌ Thinking the client app contains the AI
✔️ It just calls the agent
❌ Confusing playground with client app
✔️ Playground = testing
✔️ Client app = real usage
❌ Forgetting deployment step
✔️ Agent must be deployed first
🧠 9. Simple Memory Model
“Agent thinks, client shows”
Agent → processes and decides
Client → sends/receives messages
✅ Summary
To create a lightweight client application for an agent in AI-901:
Create and configure agent in Azure AI Foundry
Deploy the agent
Build a simple app using SDK/API
Send prompts and display responses