---
title: "How Do I Review or Export Usage?"
description: "Pick the chatbot in the sidebar's chatbot menu, then choose **Usage Log** from your account menu at the bottom of the sidebar."
category: "Quick FAQs"
canonical: "https://www.proxyai.app/help/quick-faqs/how-do-i-review-or-export-usage"
---

> For the complete documentation index, see [llms.txt](https://www.proxyai.app/llms.txt). Request public content pages with `Accept: text/markdown`.

# How Do I Review or Export Usage?

The usage log shows channel, date and time, type, input and output tokens, and cost deduction for each event. When the model provider reports that part of the input came from its prompt cache, the row shows how many tokens were cached. Filter by date range and channel, expand rows for details, move through paginated results, or choose **Export CSV**.

Dashboard summaries also show daily token use for the last 14 days and spend by product for the current month.

### Compare the same scope

Select the intended chatbot before applying filters. Choose a date range that includes the test or customer conversation you are investigating, and keep the channel filter in mind when comparing the log with a broader dashboard total. A filtered view intentionally shows less activity than an account-wide summary.

Read the event type as well as its token counts. A chatbot reply, knowledge operation and other platform activity describe different work. Where details are available, expand the row before deciding which setting caused a deduction. Prompt-cache information explains part of a model request rather than a separate customer message.

Export CSV when you need to sort or compare records outside the dashboard. Keep the selected date range, chatbot and channel with the exported file so a later reader understands what it includes. Avoid sharing customer information that is unnecessary for a billing question.

For an example of reviewing behavior alongside cost, see [Audit Cost and Behavior Together](/help/pro-tips/audit-cost-and-behavior-together). Change one setting at a time and use a consistent test conversation to make the comparison useful.
