Overview

Connect AI assistants to your HPE Aruba Networking Central instance and query live network data through conversation.

🚧

Unofficial Community Project

This is not an officially supported product of HPE. It is provided as-is, with no warranty or guarantee of fitness for any purpose. Review your organization's corporate device and data policies before connecting this server to any AI assistant.

---

The Central MCP Server wraps the Central REST APIs and exposes them as MCP (Model Context Protocol) tools. Once configured, an AI assistant in a supported client like Claude Desktop, VS Code with GitHub Copilot Chat, or Claude Code can answer network questions by calling these tools and reporting the live results.

📘

What is MCP?

The Model Context Protocol (MCP) is an open standard that defines how AI models connect to external tools and data sources. It provides a uniform interface for any MCP-compatible AI client (Claude, GitHub Copilot). These clients can connect to any MCP server without custom integration work. The Central MCP Server implements this protocol so that any supported client can call Central's APIs through the same connection setup.

Learn more at modelcontextprotocol.io.

What You Can Ask

Once connected, you can ask your AI assistant questions like:

• "Which sites have poor health scores right now?"
• "Show me all failed wireless clients at site HQ in the last 24 hours."
• "What events happened on switch SW-CORE-01 yesterday?"
• "Are there any critical alerts active on the Boston campus?"
• "Compare the health of Site A and Site B."
• "How many clients are connected at the Chicago office right now?"

The assistant retrieves live data from Central to answer each question. It does not draw from cached or static snapshots.

What It Cannot Do

The MCP server is read-only. It exposes only GET operations against the Central API. It cannot:

• Modify device configuration or apply templates
• Provision or deprovision devices
• Trigger firmware upgrades or reboots
• Create, acknowledge, or clear alerts
• Write to or modify any data in Central

📘

Note

This constraint is intentional. The server is scoped to monitoring and diagnostics use cases. Any action that changes network state must be performed directly in Central.

What the Server Exposes

Tools

The server provides 9 underlying tools, organized by resource type:

Tools are invoked automatically by the AI assistant in response to your questions. You do not call them directly.

📘

Dynamic Tool Loading

The server uses dynamic tool loading. Instead of sending the full catalog of 9 tools to your AI client upfront, which burns context tokens before you've asked a question, which gives the AI meta-tools for on-demand discovery:

• Search — finds relevant tools by keyword
• GetSchemas — fetches parameter details for tools it needs
• Execute — runs code that chains multiple tool calls in a single request

Benefits: smaller token footprint, faster responses, and better multi-step orchestration. This is transparent to you — ask questions naturally and the AI discovers the right tools on its own. You won't see central_* tool names in your client's tool picker, and that's expected.

Guided Prompts

The server also includes 10 built-in guided prompts. These are pre-built investigation workflows that instruct the AI to run a specific sequence of tool calls and return a structured analysis. These cover common monitoring tasks like site troubleshooting, failed client investigation, and network-wide health reviews.

Data and Security

All queries go directly from your machine to your Central instance via the Central REST API. No network data passes through the MCP server itself or to any third party.

Your AI assistant does receive the API responses, the same data you would see in the Central UI. Before connecting, confirm that your organization's security policies permit AI assistant access to this data.

🚧

Warning

Never type or paste credentials directly into the AI chat window. Anything typed into chat is sent to the AI provider. See How credentials stay secure for how config file credentials are protected.

Prerequisites

Before you configure the server, you need:

• A Central account with API access enabled
• uv installed on your machine
• One of the supported MCP clients like Claude Desktop, VS Code with GitHub Copilot Chat, or Claude Code

Supported Clients

The Central MCP Server works with any MCP-compatible client. The guides below cover three common setups — if you're using a different client, follow its MCP server configuration instructions and use the same uvx --prerelease=allow central-mcp-server command.