Product leadership · Network automation
Catalyst Center MCP · Product case study

Making network operations accessible to agent workflows.

I led product management for this project, and my team delivered it for customers: a Model Context Protocol (MCP) server that exposes Catalyst Center capabilities to compatible AI clients through structured tools.

Product management lead Team delivery for customers Repeatable Podman setup
The product challenge

Turn network platform capabilities into an agent-ready interface.

Catalyst Center provides network APIs. The project packages generated API tools behind an MCP server so compatible AI clients can discover and call those tools through a consistent protocol.

01

Make capabilities discoverable

Expose structured tools that an MCP client can enumerate and call, rather than requiring each client workflow to hand-build individual REST requests.

02

Keep the integration deployable

Package the server as a container and document a local Podman path, giving teams a practical way to try the integration without Docker Desktop.

03

Make validation understandable

Separate process health from controller authentication, then provide a clear read-only check to verify the end-to-end connection.

Product leadership and delivery

Move from integration concept to a customer-usable path.

My product management role connected the customer goal with a deliverable implementation and a repeatable setup experience. The team delivered the server for customers; this companion guide captures the steps to build and run it locally with Podman.

Product definition

Frame the experience around a compatible MCP client, a Catalyst Center target, and a deliberate account permission boundary.

Technical delivery

Use Cisco's standalone server, generated tool bundle, Streamable HTTP endpoint, and standard container build.

Adoption enablement

Document the steps that commonly slow down a first try: CLI PATH, Podman machine startup, branch choice, image build, TLS, and endpoint checks.

Customer value

Give teams a path to explore network data through AI-assisted workflows.

The product makes structured Catalyst Center API operations available to MCP-compatible clients. With appropriate account permissions, teams can begin with read-only inventory and health workflows, then decide which additional operations fit their governance model.

This describes the capability and intended adoption path. No customer names or quantified impact claims are published here.

Validation and transparency

Prove each layer before relying on the integration.

The deployment guide distinguishes a running HTTP process from a successful Catalyst Center login and an actual tool call.

Lab validation

Built the Cisco `release/3.2.3` source with Podman, confirmed the health and readiness endpoints, and completed an authenticated read-only inventory call.

Release compatibility

The lab controller was Catalyst Center 3.2.2; `release/3.2.3` was the closest available candidate. The successful read is a lab result, not a Cisco compatibility guarantee. Select and validate the branch for your own controller release.

Use a dedicated least-privilege account and verified TLS where possible. Disabling certificate verification is an isolated-lab workaround only. The MCP tool catalog can include state-changing operations; review permissions and each requested action.