Real-time and historical metrics for your deployed MCP servers. Understand how your server is being used, identify bottlenecks, and spot errors before users report them.
Dashboard Overview
Access analytics in the Concierge Platform after deploying your server with concierge deploy.
Metrics Available
Throughput
| Metric | Description |
|---|
| Requests/minute | Total MCP requests across all sessions |
| Active sessions | Currently connected clients |
| Peak concurrent | Highest simultaneous session count |
| Tool | Calls | Avg Time |
|---|
| search_products | 1,247 | 45ms |
| add_to_cart | 891 | 12ms |
| proceed_to_next_stage | 756 | 3ms |
| checkout | 423 | 230ms |
| apply_coupon | 198 | 18ms |
Tools with high call counts but low stage transition rates may indicate the LLM is stuck:consider improving tool descriptions or adding guardrails.
Error Rates
| Metric | Description |
|---|
| Error rate | Percentage of tool calls that returned errors |
| Error breakdown | Errors grouped by tool, error type, and stage |
| Error timeline | When errors spike, correlated with deploys |
Error Rate: 2.3% (last 24h)
Top errors:
checkout → "Cart is empty" (67 occurrences)
search_products → Timeout after 30s (12 occurrences)
add_to_cart → "Product not found" (8 occurrences)
Success Rate
Track end-to-end workflow completion:
Workflow Completion (last 7 days):
browse → cart: 89% of sessions
cart → checkout: 67% of sessions
checkout → done: 94% of sessions
Full funnel: 56% of sessions
Time-Based Usage
- Hourly/daily/weekly request volume
- Peak hours identification
- Trend lines for growth tracking
- Latency percentiles (p50, p95, p99) over time
Stage Analytics
See how users flow through your staged workflows:
Identify where users drop off and optimize those stages.
Alerts
Set up alerts in the platform for:
- Error rate exceeding a threshold
- Response time spikes
- Unusual traffic patterns
- Stage drop-off rate changes
