Skip to main content
Data Infrastructure Insights

DII MCP Server

Contributors netapp-alavoie
PDFs

Connect AI assistants/agents to Data Infrastructure Insights (DII) so they can investigate infrastructure inventory, telemetry, alerts, and collection health using natural language.

Note The DII MCP is a Preview feature and is therefore subject to change.

The DII MCP Server implements the Model Context Protocol (MCP) and exposes tenant-scoped tools to supported AI clients. The current product toolset is read-only: it retrieves and analyzes DII data but does not modify infrastructure or DII resources.

What you can investigate

  • Infrastructure inventory: storage, virtualization, Kubernetes, ONTAP, Azure NetApp Files, StorageGRID, and related resources.

  • Performance and capacity: object metrics, grouped aggregates, and time-series behavior.

  • Logs and events: discover log types, search events, and build event-count histograms.

  • Alerting: inspect active alerts, monitor definitions, severities, and corrective guidance.

  • Collection health: review acquisition units, data sources, supported collector types, and Kubernetes collector versions.

  • Notification posture: inspect maintenance windows, notification rules, and webhook target types.

Choose an entry point

Connect an AI client

Follow Set up the DII MCP Server for Cursor, Claude Code, VS Code, Codex, or another MCP-compatible client.

Browse available tools

Use the tool reference to understand all current tools, required discovery calls, query limits, and user-facing names.

Start with an operational workflow

Use a recipe when you already know the outcome you need:

  • Inventory collection health

  • Assess storage health

  • Identify capacity risk

  • Check Kubernetes collector currency

  • Troubleshoot performance

  • Review notification posture

How DII MCP queries work

Many DII resources have dynamic schemas. A reliable investigation follows a discovery-first sequence:

  1. List the available object or log types.

  2. Retrieve metadata for the selected type.

  3. Use only valid attributes and metrics from that metadata.

  4. Query a bounded time range.

  5. Refine the result with filters, grouping, sorting, or pagination.

See Concepts and data model for details.

Important limitations

  • All current product tools are read-only.

  • Object, performance, and log queries enforce bounded time windows. Use windows of 30 days or less unless a tool page explicitly documents a smaller limit.

  • Query fields must match the metadata returned for the selected resource type.

  • Large results are paginated or capped.

  • Tool availability and returned data depend on the tenant's collectors, enabled features, and user permissions.

  • Do not request or reproduce complete webhook endpoint addresses. See Security and privacy.

Next steps

  1. Connect your client.

  2. Try the environment inventory recipe.

  3. Review Troubleshooting if a query is rejected or returns unexpected results.