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

# Token usage

> See how many tokens your reviews use and what they cost, by day, pull request, review run, developer, model and task.

**Organization → Token usage** shows the tokens your reviews consume and an estimate of what they cost.

Owners, Billing Managers and Repo Admins can open the dashboard or query usage with `kodus usage tokens` after signing in.

For a monthly budget with alerts, see [Spend limit](https://docs.kodus.io/en/how_to_use/spend-limit.md), which links here for the full breakdown.

To check token usage and estimated cost from the terminal or export them as JSON, use [`kodus usage tokens`](https://docs.kodus.io/en/how_to_use/cli/commands#usage-tokens).

## Filters

* **Models**: tick the models to include. All are selected by default.
* **Repository**: shown when you have connected repositories.
* **View**: **Daily**, **By PR**, **By Review** or **By Developer**. **By PR** adds a **PR Number** box, and **By Developer** adds a developer picker.
* **Date range**: **Last week**, **Last 15 days**, **Last month**, **Last 3 months** or custom dates. The default is the last 15 days.

## Summary cards

Totals for the range and filters, with the change against the previous period of the same length where shown.

| Card | What it shows |
| - | - |
| **Total cost** | Estimated cost, with the average per day and per PR |
| **Total tokens** | All tokens used, including reasoning tokens |
| **Uncached input** | Input tokens billed at the full input price |
| **Cache read** | Input tokens served from the provider's prompt cache, billed at a discount, with what that saved |
| **Output** | Output tokens, including reasoning |
| **Cache write** | Input tokens that filled a cache entry. Shown only when there are some |

Cost is estimated from each model's per-token prices. Models with **Price unavailable** still contribute to token counts. The cost totals omit their spend, so treat those totals as partial estimates.

If you set a monthly budget, a **Monthly spend limit (BYOK)** bar shows this month's spend against it. It always covers the current month and ignores the date range.

## Chart

The chart follows the **View** you picked, and **Tokens** / **USD** switches its unit. **By PR**, **By Review** and **By Developer** show the 24 biggest consumers by tokens:

* **Daily**: **Usage over time**, split by uncached input, cache, output and reasoning.
* **By PR**: **Top pull requests**, the PRs that used the most tokens.
* **By Review**: **Top review runs**. Each bar is one review run, so a PR reviewed twice appears twice. A **Review activity** table below lists each run with its PR, start time, models, tokens and cost, most expensive first, up to 100 runs. Narrow the filters to see the rest.
* **By Developer**: **Top developers**, with usage attributed to the PR author.

## Where tokens go

**Where tokens go** breaks spend down by task, such as Code Review, Kody Rules Review, PR Summary, Conversation and Custom Agents, with a bar showing the mix of input, cache, output and reasoning. Expand a task to see the steps it ran and which models ran it and when. Spend that can't be tied to a task is listed as **Unrouted**.

Below it, each model has its own row with its cost. Expand a row for its tokens and cost per token type, and hover the price icon for the model's rates.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.