Service Observability MCP Server

The Service Observability (SO) MCP server enables MCP-compatible AI clients to query Virtana Service Observability using natural language. Model Context Protocol (MCP) is an open standard that connects AI clients to external data sources and tools. With the SO MCP server, you ask an AI client plain-language questions about your monitored infrastructure, such as which resources exist, how a metric is trending, or whether an entity has active alerts, without needing to know the underlying Service Observability APIs.

About the Service Observability MCP server

The SO MCP server connects an AI client to Virtana SO and answers questions using your monitored data. Describe what you want in plain language, and the server retrieves the matching entities, metrics, anomalies, or events.

With the SO MCP server, you can ask an AI client to do the following:

Discover the entity types modeled in your environment.
Search for monitored entities by name, type, or property.
Inspect the fields and metrics available for a type or entity.
Retrieve metric summaries and trends over a time range.
Analyze a metric across many entities at once.
Investigate the root cause of a service issue or impact.
Detect anomalies and review events and alerts.

Virtana SO is a SaaS offering, and the SO MCP server uses data from the SO SaaS platform. You can connect to the SO MCP server from a local or external MCP-compatible client, but the monitored data always comes from the SaaS environment. Service Observability isn't offered as a standalone on-premises deployment for this capability.

Ways to access the Service Observability MCP server

You can use the SO MCP server in two ways. Choose the method that matches where you want to work.

Built-in Copilot

The Virtana Copilot in SO already uses the SO MCP server. Wherever the built-in Copilot chatbot is available, you can ask questions about your SO data directly, with no separate client setup. This is the simplest way to get started.

External MCP-compatible client

You can connect an external AI client, for example, Goose, Claude Desktop, or Cursor, to the SO MCP server over HTTP. This method requires a one-time setup. Include the SO MCP endpoint URL and an API key in the external AI client while configuring. Use this method when you want to work in your own AI client or integrate Service Observability data into an existing workflow.

Both methods use natural-language prompts and provide the same query capabilities. The difference is where you work and whether you configure a client yourself.

Set up an external MCP client

Generate a connection URL and API key in Service Observability, and then add them to your client to use the SO MCP server from an external AI client.

Before you begin

Before you connect an external MCP client, make sure you have the following:

A Virtana Service Observability account with valid login credentials. If you do not yet have an account, contact your Virtana administrator. For more information on logging in, see Logging in to Service Observability.
An AI client that supports MCP over HTTP and allows custom request headers.
Permission to create API clients in the SO administration interface. For details on managing API access in Service Observability, see API clients and keys.

Generate an endpoint URL and API key

The MCP Endpoint URL is the network address that an external MCP client uses to connect to the SO MCP Server. The client cannot locate or communicate with the server without it.

Determine your endpoint URL: Use the API Address of your SO instance and append /mcp to the end of the URL. The resulting URL is the MCP server endpoint that you will configure in your MCP client.
Generate an API key: Generate an API key as described in Generate a key.
Important
Copy the generated API key immediately and store it securely. Treat the key like a password, as anyone with the key and the endpoint URL can query your SO data.

You will need both the API key and the MCP server URL to configure your client.

Configure your MCP client

When you add the SO MCP server to an MCP-compatible client, use the following settings. For more information, see Configure extensions.

Setting	Value
Transport	Streamable HTTP
Server URL	Your Service Observability endpoint URL with `/mcp` appended
Header name	`zenoss-api-key`
Header value	Your generated API key

After the client connects successfully, you can start asking questions in natural language.

Note

Each MCP client has its own configuration flow and may ask for an extension name, a model provider, or other settings. The values specific to Service Observability are the MCP server URL and the zenoss-api-key header. Your AI client also needs its own model provider (LLM) configured before it can process prompts.

In this section: