Query Your Production Data with Natural Language

Stop switching between your AI assistant and Honeycomb's web interface. This MCP server connects your LLM directly to your observability data, letting you debug production issues and analyze system behavior through conversation.

Why You Need This

You're already context-switching between your code editor, monitoring dashboards, and AI chat. Every time you need to check error rates, analyze slow queries, or investigate an incident, you break flow to open another browser tab, remember filter syntax, and manually correlate data points.

The Honeycomb MCP server eliminates that friction. Ask your AI assistant "What's the P95 latency for the payment service in the last hour?" and get immediate answers from your actual production data.

What You Get

Direct Data Access: Query any Honeycomb dataset through natural language. Your AI assistant becomes a first-class interface to your observability stack.

Multi-Environment Support: Seamlessly switch between production, staging, and development environments without changing tools or context.

Rich Analytics: Run complex queries with breakdowns, filters, percentiles, and time-based analysis. The same power as Honeycomb's query builder, but through conversation.

SLO and Alert Monitoring: Check SLO burn rates, trigger status, and alert configurations without leaving your development environment.

Intelligent Caching: Built-in TTL-based caching reduces API calls and improves response times for metadata queries.

Real-World Usage

Incident Response: "Show me error rates by service for the last 30 minutes" → immediate breakdown without dashboard navigation

Performance Investigation: "What's the slowest endpoint in production right now?" → sorted P95 latencies with actual values

Deployment Validation: "Compare API response times before and after the 2 PM deploy" → side-by-side analysis

Capacity Planning: "Show me database connection pool utilization trends over the past week" → historical analysis with breakdowns

Integration That Just Works

Compatible with the MCP clients you're already using:

• Claude Desktop for general development
• Claude Code for coding sessions
• Cursor for AI-powered development
• Windsurf for collaborative work
• Goose for automated workflows

Configuration is straightforward - add your Honeycomb API key to your MCP client's environment variables and start querying.

The Technical Details

Built specifically for Honeycomb Enterprise customers with full API access. The server runs locally on your machine using Node.js 18+, communicating with your MCP client over STDIO.

Query Capabilities:

• All calculation types (COUNT, P95, HEATMAP, etc.)
• Complex filtering with multiple operators
• Multi-column breakdowns and custom sorting
• Absolute and relative time ranges
• Statistical analysis with automatic summary generation

Resource Management:

• Optimized responses to minimize context window usage
• Configurable caching for different resource types
• Support for EU endpoints and custom API configurations
• Multi-environment API key management

Enterprise Features:

• SLO monitoring and budget tracking
• Trigger analysis and notification management
• Dataset schema exploration
• Trace link generation for deep investigation

Get Started Now

git clone https://github.com/honeycombio/honeycomb-mcp.git
cd honeycomb-mcp
pnpm install && pnpm run build

Add to your MCP client configuration:

{
  "mcpServers": {
    "honeycomb": {
      "command": "node",
      "args": ["/path/to/honeycomb-mcp/build/index.mjs"],
      "env": {
        "HONEYCOMB_API_KEY": "your_api_key"
      }
    }
  }
}

Start asking your AI assistant about your production systems instead of hunting through dashboards. Your observability data becomes as accessible as your codebase.