Usage Examples¶
This section provides example queries you can ask your MCP client (GitHub Copilot, Claude Desktop, etc.) when using the Harvest MCP Server. For more examples and community discussions about MCP usage, see: Harvest MCP Discussion
Higher-capability language models provide better analysis and insights. When possible use the latest model versions with larger context windows.
Getting the Best Results¶
Use Harvest MCP Provided Prompt¶
For optimal analysis and insights, always start by setting the Analysis Expert prompt available in the Harvest MCP server. This prompt provides your MCP client with instructions for analyzing Harvest data.
Access the prompt: Most MCP clients support prompts - look for a /mcp command or prompts menu to select the "Analysis Expert" prompt.
Reference Questions¶
Below are example questions that work well with the Harvest MCP Server:
Infrastructure Health¶
"What's the overall health of my infrastructure?"
Expected response: Comprehensive health summary showing cluster status, active alerts, capacity issues, and immediate action items with priority levels.
Capacity Analysis¶
"Which volumes are approaching capacity limits and need attention?"
Expected response: List of volumes above 90% utilization with details about clusters, SVMs, growth trends, and recommended actions including timeline for capacity expansion.
Performance Investigation¶
"Show me systems experiencing performance issues with high latency or IOPS"
Expected response: Analysis of volume and node performance metrics, identification of hotspots, correlation with capacity utilization, and specific recommendations for optimization.
Trend Analysis¶
"Analyze storage growth patterns across my clusters over the past 30 days and predict when I'll need to add capacity"
Expected response: Detailed growth analysis by cluster and aggregate, mathematical projections of space consumption, identification of fastest-growing workloads, and recommended expansion timeline with sizing guidance.
Integration Tips¶
- Always set the prompt first before asking questions
- Use specific questions rather than vague requests
- Ask follow-up questions to dive deeper into specific areas
- Combine multiple areas (e.g., "Show me capacity and performance issues together")
- Request different perspectives (executive summary vs. technical details)
MCP Clients¶
Common MCP clients that work with Harvest MCP Server:
- GitHub Copilot: Integrated in VS Code, supports MCP connections
- Claude Desktop: Anthropic's desktop application with MCP support
- Custom MCP Clients: Any application implementing the MCP standard