---
title: "AI Services Usage Statistics: Costs, Tokens, Trends | Hostman Docs"
description: "Track AI Gateway and agent usage in the Hostman dashboard: costs, requests, tokens, cache hit rate, top models, and trends for the last 90 days."
---

> For the complete documentation index for AI agents, see [llms.txt](https://hostman.com/llms.txt).

The [Activity](https://hostman.com/my/cloud-ai/activity) page of your dashboard shows usage across all your [AI Gateway API keys](https://hostman.com/docs/ai-services/ai-gateway/) and [AI agents](https://hostman.com/docs/ai-services/ai-agents/) for a selected period. You can see how much you spend, how many requests and tokens you use, how well caching works, and which models, keys, and agents drive your usage.

The page has two views:

-   **Activity**: what happened during the period. Cost, requests, tokens, caching, top keys and agents, and a breakdown by model.
-   **Trends**: what changed. Models, API keys, and agents whose usage has grown noticeably.

## Limitations

-   Data is available for the last 90 days. Older data cannot be retrieved.
-   Days follow your browser’s time zone. Each daily bar covers requests from 00:00 to 24:00 local time.

## View Usage for a Period

1.  In the dashboard, go to **AI Services** → **Activity**.
2.  Select **Show activity**.
3.  Select a period or set a custom date range.

The page shows these cards:

| **Card** | **Description** |
| --- | --- |
| Spend | Total charges for API key and agent requests |
| Requests | Number of requests sent to models |
| Tokens | Total input and output tokens. Reasoning tokens are counted as output tokens |
| Cache share | Share of input tokens served from cache. The higher the rate, the lower the cost of repeated requests |
| Avg. price per 1M tokens | Average cost of one million tokens across all models for the period |

Each card includes a daily chart for the selected period. Hover over the chart to see details. If a card has no chart, the value didn’t change during the period.

#### Compare With the Previous Period

Each card shows how its value changed compared to the previous period of the same length. For example, if you select the 1st through the 15th of the current month, the page compares it with the 15 days before the 1st.

-   Cost changes are shown as an amount and a percentage. Other metrics are shown as a percentage only.
-   If the value in the previous period was zero, a percentage can’t be calculated, so only the absolute difference is shown.
-   Percentages are relative to the previous value. If the cache hit rate grows from 30% to 33%, the change is 10%, not 3%.

## Find Cost Drivers

The **Spend** card shows your total cost and how it changed. 

To see what makes up the total, check the following:

-   **Top API keys** and **Top agents**: the keys and agents with the highest charges. Each entry shows the cost and the number of tokens.
-   **Cost by model**: which models cost the most and on which days your costs peaked. The three most expensive models have their own colors, and the rest are grouped as **Other**.

If you delete an API key, it stays in the list under its original name. A deleted agent appears as Deleted agent.

## Analyze Token Usage

The **Tokens** card shows the total number of tokens. To see what they consist of, see the **Tokens by type** chart below. It splits tokens into input, output, and reasoning tokens, so you can tell whether usage comes from large prompts, long responses, or model reasoning.

If **Avg. price per 1M tokens** grows while the token volume stays stable, your workload has probably shifted to more expensive models.

## Check Caching Efficiency

The **Cache share** card shows what share of input tokens was read from cache. The **Input caching** chart below breaks this down by day, showing cached and uncached input tokens.

Cached tokens are billed at a lower rate than regular input tokens, so the cache hit rate directly affects your cost. The higher the rate, the cheaper repeated requests are.

## Compare Models by Cost and Requests

Use **Requests by model** to see which models are used the most, and compare it with **Cost by model** to see which ones cost the most. The three leading models are highlighted in both charts, and the rest are grouped as **Other**.

## Track Growth With Trends

The **Trends** view  shows where cost, tokens, or requests grow the fastest. It includes three charts: models, API keys, and agents.

1.  Go to **AI Services** → **Activity**.
2.  Select **Show trends**.
3.  Select a period or set a custom date range.
4.  On each chart, select a metric: **Cost**, **Tokens**, or **Requests**.

To the right of each chart, you’ll see the three fastest-growing items for the selected metric. New items that had no usage in the previous period come first, followed by the rest in descending order of growth relative to the previous period.

Minor fluctuations are filtered out. An item appears in the list only if it grew by at least 10% and accounts for at least 1% of the total on the chart.

The list can be empty for two reasons:

-   There is no notable growth, meaning no significant changes were found.
-   There is nothing to compare with: the previous period falls outside the 90-day window, or there was no usage in it.

## Filter by API Key, Agent, or Model

You can filter the statistics to see data only for specific API keys, agents, or models. The filter applies to everything on the page at once, including trends.

1.  Go to **AI Services** → **Activity**.
2.  Click the filter icon in the upper-right corner.
3.  Set the filter parameters:
    -   **Source**: show all sources, only agents, or only API keys.
    -   **Keys**, **Agents**, **Models**: select specific items. You can select up to 50 in each list.
4.  Click **Apply**.

The filter lists only items that were active during the selected period. If an API key had no requests, it won’t appear in the list.