> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getgriffinapp.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Metrics

> Track success rates, latency, and uptime

Griffin aggregates run results into metrics so you can track the health of your APIs over time.

## Viewing metrics

```bash theme={null}
# Summary across all monitors (last 24 hours)
griffin metrics

# Metrics for a specific environment
griffin metrics production

# Different time windows
griffin metrics --period 1h    # Last hour
griffin metrics --period 6h    # Last 6 hours
griffin metrics --period 24h   # Last 24 hours (default)
griffin metrics --period 7d    # Last 7 days
griffin metrics --period 30d   # Last 30 days

# JSON output for scripting
griffin metrics --json
```

## Available metrics

### Summary metrics

| Metric             | Description                          |
| ------------------ | ------------------------------------ |
| **Total monitors** | Number of deployed monitors          |
| **Passing**        | Monitors with recent successful runs |
| **Failing**        | Monitors with recent failures        |
| **Total runs**     | Number of runs in the time period    |
| **Success rate**   | Percentage of successful runs        |
| **Uptime**         | Overall availability percentage      |

### Latency metrics

| Metric  | Description                   |
| ------- | ----------------------------- |
| **p50** | Median response time          |
| **p95** | 95th percentile response time |
| **p99** | 99th percentile response time |

### Per-monitor metrics

The hub also tracks detailed metrics per monitor:

* Run count, success count, failure count
* Min, avg, and percentile latencies
* Last run time and status
* Recent failure details
* Error distribution

## How metrics are calculated

The hub aggregates run results into hourly and daily buckets. When you query metrics, the API rolls up the relevant buckets for your time period.

This means metrics are always based on actual run data — there's no sampling or estimation.