Connect the AI assistant to SolarWinds Observability Self-Hosted with the MCP server

The Model Context Protocol (MCP) provides an open, standardized framework that connects AI models to data sources and execution tools in SolarWinds Observability Self-Hosted.

The SolarWinds Observability Self-Hosted server lets theAI assistant and other MCP-compatible AI clients query your monitoring data through a standard set of tools. This topic explains how to access the MCP server, authenticate, and configure a client.

Key capabilities

Secure context retrieval: The SolarWinds Observability Self-Hosted MCP server enables the AI assistant and supported external AI tools to query monitoring telemetry, entity configurations, alerts, and metrics.
Permission-aware tool execution: MCP operations run under the authenticated user session, honoring existing account boundaries, entity limitations, and node view restrictions.

Before you begin

Upgrade to SolarWinds Observability Self-Hosted 2026.4 or later.
You need to have the AI assistant configured and Platform Connect enabled. See AI assistant.
Use an online environment.
Choose an MCP-compatible AI client that supports the Streamable HTTP transport and lets you set a basic Authorization header.
Make sure the client can reach your main polling engine over HTTPS.

If your SolarWinds Platform Web Console uses an HTTPS certificate, ensure that LLM clients trust the certificate. The AI assistant uses the same endpoint as the SolarWinds Platform Web Console, including the certificate if HTTPS is enabled.
Create or choose an account for the connection. See Create users in the SolarWinds Platform.

SolarWinds recommends a dedicated account with the minimum permissions required.

Access the MCP server

Use the following MCP endpoint:

https://<FQDN>/mcp

Replace <FQDN> with the fully qualified domain name you use to access your SolarWinds Observability Self-Hosted main polling engine.

The MCP service runs only on the main polling engine. Additional polling engines (APEs) do not host the service. Always point your client to the main polling engine, including in multi-engine deployments.

Authenticate

The MCP server supports HTTP Basic authentication.

Send your SolarWinds Observability Self-Hosted username and password in the Authorization header of every request, using the following format:

Authorization: Basic <base64(username:password)>

Requests run as the SolarWinds Observability Self-Hosted account you use to authenticate. Results reflect that account's permissions, so users can access only the data they are already allowed to see in SolarWinds Observability Self-Hosted.

Configure your AI assistant

Configuration formats vary by client. The following example shows a remote server URL and a custom authentication header:

{
  "mcpServers": {
    "osh": {
      "type": "http",
      "url": "https://<FQDN>/mcp",
      "headers": {
        "Authorization": "Basic <base64(username:password)>"
      }
    }
  }
}

Replace <FQDN> with your main polling engine's fully qualified domain name.

Replace <base64(username:password)> with the Base64-encoded value of your SolarWinds Observability Self-Hosted username and password, separated by a colon.

Refer to your client's documentation for the exact configuration syntax.

Enable/disable the MCP server

In the SolarWinds Platform Web Console, click the Platform Connect icon () in the menu bar. The Platform Connect details are displayed.
Enable or disable the MCP server toggle.

Things to know

The MCP server is available only for online environments in this release.
Data returned by the tools is sent to the AI client and the model provider you choose. Review your organization's data-handling and AI usage policies before connecting a client.
Access to the AI assistant in the SolarWinds Platform Web Console is controlled separately, through the AI Settings section of user and group permissions.

Search SolarWinds Support