AI-Driven Network Operations with Catalyst Center and MCP

This lab gives you hands-on experience using an AI assistant to query and troubleshoot a pre-configured Cisco SD-Access fabric managed by Catalyst Center, using natural language instead of the GUI or CLI. 

1. You'll start by exploring how the assistant is built. It combines an ambient cache of network data with on-demand MCP (Model Context Protocol) tool calls that reach Catalyst Center and live devices. 
2. Work through six hands-on scenarios that put real questions to the AI assistant about your own live SD-Access fabric, so the answers, and what you learn from them, reflect the current state of your lab and not just what a model happens to know.
3. You'll then put it to work in a hands-on challenge: an on-call engineer is about to open a change window on an access switch and needs to know the blast radius if one of its interfaces goes out of service. With no step-by-step prompts, you'll use only the assistant to find which clients are affected, which virtual network they belong to, and whether BGP peering on the upstream router stays up. 

By the end of this experience, you'll understand how an AI assistant gets its answers from live network state, how to prompt it effectively, and how to tell evidence-backed answers from plausible guesses, so you can judge for yourself when to trust what it tells you.

Goals

1. Understand how an AI network assistant gets its answers, from cached network data and from live MCP tool calls to Catalyst Center and devices.
2. Use natural-language prompts to investigate a real operational question in an SD-Access fabric.
3. Build the habit of verifying AI answers against tool evidence before acting on them.

After completing this lab, you will be able to:

1. Explain how the AI Lab Assistant uses cached data and on-demand MCP tool calls.
2. Query interface status, connected clients, virtual network membership, and BGP state in natural language.
3. Treat "data unavailable" as a valid answer instead of filling the gap with a guess.
4. Support a change-impact assessment with evidence from the assistant.

• Cisco Catalyst Center Appliance (3.2.3)
• 4x Catalyst 9K (IOS-XE 17.18)
• 3x C8000v  (IOS-XE 17.9.3a)
• Cisco Identity Services Engine (3.5 Patch 3)
• 4x Windows 11 Clients
• 1x Vyos Router
• Ubuntu MCP Server (llama-3.3-70b)