kanmi-perf-mcp

v0.1.0 · MIT License (Free)

Model Context Protocol server for AI assistant integration with KanmiPerf performance data. Enables Claude Code and other MCP clients to access real-time performance metrics.

Install

npm install -g kanmi-perf-mcp

Start the Server

# Default (http://127.0.0.1:3960)
kanmi-perf-mcp

# Custom port/host
kanmi-perf-mcp --port 4000 --host 0.0.0.0

Claude Code Configuration

{
  "mcpServers": {
    "kanmi-perf": {
      "command": "node",
      "args": ["path/to/kanmi-perf-mcp/src/server.js"],
      "env": {
        "PORT": "3960"
      }
    }
  }
}

Available Resources

Core Web Vitals — LCP, CLS, INP, FCP, TTFB
Runtime Metrics — Long tasks, memory usage, network info
Historical Data — Performance trends over time
Site Comparisons — Relative performance data

MCP Tools

Tool	Description
`kanmi_analyze_site`	Deep performance analysis
`kanmi_get_recommendations`	AI-powered optimization suggestions
`kanmi_compare_sites`	Relative performance comparison
`kanmi_detect_regressions`	Performance degradation detection
`kanmi_export_data`	Export performance reports

API Endpoints

Resources

GET /mcp/resources              # List all data sources
GET /mcp/resources/site-{id}    # Get specific site data
GET /mcp/resources/sites        # Get site list

Analysis

POST /mcp/tools/analyze         # Analyze site performance
POST /mcp/tools/recommend       # Get recommendations

Real-time

GET /events                     # SSE stream for live updates

Configuration

Variable	Description	Default
`PORT`	Server port	3960
`HOST`	Server host	127.0.0.1

Requirements

Node.js 18+
At least one KanmiPerf package for data source
MCP-compatible AI assistant (Claude Code recommended)

Repository

github.com/konfirmed/kanmi-perf