npm install -g @reaatech/mcp-load-test-cli
# or
pnpm add -g @reaatech/mcp-load-test-cli

# Auto-detect transport, ramp from 1→50 over 60s
mcp-load-test load --endpoint https://api.example.com/mcp
 
# Soak at constant 25 concurrency for 30 minutes
mcp-load-test soak --endpoint https://api.example.com/mcp --concurrency 25
 
# Spike test: 10 baseline, burst to 100, repeat 3×
mcp-load-test spike --endpoint https://api.example.com/mcp --baseline 10 --spike 100
 
# Markdown output to file with breaking point detection
mcp-load-test ramp --endpoint https://api.example.com/mcp --max-concurrency 100 --breaking-point --format markdown --output report.md
 
# Compare two reports
mcp-load-test compare --baseline baseline.json --current current.json

# load-test-config.yaml
endpoint: "https://api.example.com/mcp"
transport: "http"
auth:
  bearerToken: "${MCP_BEARER_TOKEN}"
 
profile:
  type: "ramp"
  minConcurrency: 1
  maxConcurrency: 100
  rampDurationMs: 300000
  holdDurationMs: 120000
 
patterns:
  - name: "explore-then-act"
    weight: 0.7
  - name: "multi-step-workflow"
    weight: 0.3
 
breakingPoint:
  enabled: true
  thresholds:
    errorRate: 0.05
    latencyP99: 5000
 
output:
  format: "markdown"
  file: "./load-test-report.md"