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

# Usage

> This month's consumption across the whole account — calls, bytes, tokens and cost.

The usage meters on `/settings/billing` cover the **whole account**, not one MCP, for the current month.

| Figure            |                                                      |
| ----------------- | ---------------------------------------------------- |
| MCPs              | How many tracked servers exist                       |
| Members           | How many people are in the account, invited included |
| Calls this month  | Tool calls across every MCP                          |
| Bytes this month  | Response bytes across every MCP                      |
| Tokens this month | Those bytes as tokens, `bytes / 4`                   |
| Cost this month   | Those tokens priced at \$3 per million               |
| Tools tracked     | Registered tools across every MCP                    |

## What the cost figure means

The same thing it means everywhere else in MCPulse: **what your servers cost the people running them**, in context, this month. It is not what MCPulse charges you — that is your [plan](/billing/plans), and it does not move with your traffic.

It uses the same two constants as [cost per session](/metrics/cost), from the same service, so an account-wide figure and a per-MCP figure cannot disagree about arithmetic.

## The month is a UTC month

Like every other date boundary in the product. See [Date ranges](/api/concepts/date-ranges).

## The calls meter is the one that bites

**Calls this month** is metered against your [plan](/billing/plans), and reaching it stops recording: ingest answers `402`, a banner appears above the overview, this meter turns red, and the chart is flat from the moment it stopped.

Your server is unaffected. The SDK treats any non-2xx as a failed batch and drops it, exactly as it treats a network failure, so a cap costs you data and never a tool call.

The verdict is cached for a minute, the same way an API key is — a database round trip per batch is a round trip inside somebody's tool call. So going over, or upgrading, takes up to a minute to take effect.

<Note>
  **Sample data is excluded.** The `Acme Orders` MCP every new account is seeded with carries thousands of fabricated calls; counting them would open a brand-new account a third of the way through its allowance having made no calls at all.
</Note>

## The other limits

|                       |                                                                  |
| --------------------- | ---------------------------------------------------------------- |
| **MCPs**, **Members** | Refused at creation, with the plan named in the message          |
| **History**           | Metric routes clamp the range rather than refusing it            |
| **Alerts**            | Creating, re-enabling and digest subscription are closed on Free |

Separately from all of these there is an ingest **rate limit** — 10,000 payloads per minute per key — which protects the API rather than metering a plan. See [Rate limits](/api/concepts/rate-limits).

## Related

* [Plans](/billing/plans)
* [Cost per session](/metrics/cost)