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

# Actor metrics

> Request, latency, resource, and stored-state aggregates for a StatefulActor type.

> **Requires CLI v0.5.8 or newer.**

`telnyx-edge actors metrics <type>` summarizes a StatefulActor type's recent request and resource metrics, defaulting to the last 24 hours:

* request count, with success and error rates
* request latency (p50/p95/p99/average)
* average request/response byte sizes
* CPU and memory usage against limits
* stored state size — the type's total `ctx.storage.sql` size across all its instances over the window (per-instance point-in-time size stays on `actors instances`)

```bash theme={null}
telnyx-edge actors metrics Counter
telnyx-edge actors metrics Counter --since 6h --json
```

## Reading the output

An actor's requests are its method calls (auto-RPC, `fetch()`, `alarm()`) — the same traffic the [invocation logs](/docs/edge-compute/stateful-actors/observability/logs#invocation-logs) record one line at a time. Outcomes are `ok`/`error`, not HTTP status classes, so there is no client-error split like function metrics have.

## Flags

| Flag | Description |
| - | - |
| `--since` | How far back to summarize (e.g. `1h`, `24h`, `90h`; default `24h`, max `90h`) |
| `--json` | Output the raw metric aggregate response as JSON |

`--since` is capped at `90h`: past the backend's invocation-window budget, the request, latency, and byte-size fields would come back null while CPU, memory, and storage kept reporting, so the CLI refuses wider windows rather than return a partial response. (The root `metrics <function>` command keeps its `168h` max — the shorter window is actor-specific.)

Full command reference: [CLI reference](/docs/edge-compute/reference/cli#actors-metrics).