Mannequin Context Protocol (MCP) servers present a brand new strategy to unify automation and observability throughout hybrid Cisco environments. They allow an AI consumer to routinely uncover and use instruments throughout a number of Catalyst Heart clusters and Meraki organizations.
In the event you’re interested by how this works, now’s the time to see it in motion.
On this new demo, Cisco Principal Technical Advertising and marketing Engineer Gabi Zapodeanu reveals how a single AI consumer routes natural-language queries to the fitting device, retrieves responses from a number of domains, and helps you troubleshoot or report in your community extra effectively.
See MCP in Motion: Catalyst Heart and Meraki Integration
Within the video under, Gabi demonstrates how MCP servers allow an AI consumer to work together with instruments throughout a number of platforms. You’ll be taught:
- How the consumer connects to a number of MCP servers and discovers accessible instruments.
- How these instruments are chosen and executed in actual time based mostly on person intent.
- How a single question can span clusters and organizations utilizing patterns like cluster = all.
The video consists of sensible walkthroughs of multi-cluster stock lookups, problem correlation throughout, and a BGP troubleshooting workflow constructed from fundamental instruments.
Understanding MCP Structure and Workflow
MCP makes use of a client-server protocol that permits an AI assistant to connect with a number of MCP servers and dynamically uncover accessible device definitions. Here’s what the total workflow appears to be like like:
- An AI consumer, powered by a big language mannequin, connects to a number of MCP servers.
- Every server offers a listing of instruments—both prebuilt runbooks or auto-generated APIs.
- A person asks a query; the AI consumer selects the suitable device, fills within the parameters, and sends the request.
- The instruments execute, return information, and the AI responds to the person.
This allows asking a single query—akin to “The place is that this consumer linked?”—and receiving solutions from a number of clusters and organizations.
Crucial Instruments vs. Declarative Instruments in MCP Servers
The demo explains two kinds of instruments supported by MCP servers:
- Crucial instruments are predefined sequences written in Ansible, Terraform, or Python. They’re greatest suited to write duties the place guardrails and strict execution order are essential.
- Declarative instruments are auto-generated from YAML information and are perfect for read-heavy duties akin to stock, occasion lookup, or compliance checks. Additionally they help pagination with offset and restrict parameters.
Gabi shares examples of each sorts, demonstrating their use in actual situations like firmware checks and cross-domain consumer discovery.
Troubleshooting and Compliance Utilizing Generative AI Flows
Past single-tool calls, MCP helps multi-step workflows. These generative AI flows allow you to:
- Correlate occasions
- Determine root causes of points akin to BGP flaps
- Run compliance checks or acquire telemetry throughout websites
- Apply guardrails for modifications, making certain solely trusted runbooks are used for configuration actions
The MCP consumer learns from device utilization patterns and may recommend new instruments based mostly on frequent API calls.
How you can Get Began and What’s Subsequent
This demo offers a transparent, sensible introduction to MCP for anybody working in NetOps or DevOps. You’ll acquire a greater understanding of:
- Why MCP issues right this moment
- How you can join MCP to your Cisco platforms
- The kinds of instruments and workflows it helps
- How you can construction your personal instruments utilizing YAML or SDKs
Watch the total replay:
Subscribe to Cisco DevNet on YouTube to get notified when new demos go reside.
