---
title: "Agent Analytics"
id: "1803"
type: "page"
slug: "analytics"
published_at: "2026-09-20T21:51:54+00:00"
modified_at: "2026-09-21T01:36:28+00:00"
url: "https://xedant.com/agents/data/docs/analytics"
markdown_url: "https://xedant.com/agents/data/docs/analytics.md"
excerpt: "The Analytics page shows how the agent works for you: how many tokens (the units…"
---

# Agent Analytics

[https://xedant.com/agents/data/docs/analytics.md](https://xedant.com/agents/data/docs/analytics.md)

The Analytics page shows how the agent works for you: how many tokens (the units of text an AI model “thinks” in) and how much money it uses, when the load peaks, which models run. It is cost control in plain view: you see whether automating data collection pays off and where to save.

## Chats overview

A table of all chats, sortable by any column: title, date, message count, tokens, cost, duration, messages per second and model. Handy for finding the most expensive or longest conversations and understanding where the resources go.

## Metrics

- messages — how many tasks and questions were sent;
- tokens — input (what the agent read), output (what it wrote) and tokens from the temporary storage where the agent keeps ready pieces of its answers;
- cost — what it costs in money;
- messages per second and duration — how fast the agent responds.

## Distributions

Cost, tokens, duration and messages per second are shown as distributions: you can see where most tasks are cheap and fast, and the rare ones are expensive and slow.

## Load over time

Hours of the day and days of the week — messages, response speed and cost. You can see when the agent is busiest, to move heavy tasks to quiet hours or, the other way round, to keep cheap models serving the peaks.

## By model

Cost, tokens and speed broken down by model. This is a choice about efficiency and price: if routine tasks (checking exports, small fixes) run on the most expensive model, move them to a cheap one — quality will not suffer, and the bill drops noticeably.

## Date ranges

The data can be viewed for 7, 14, 30, 90 or 365 days — or all time. Narrowing the overview to the period you need is easy, for example to evaluate the month after rollout.

## Bulk collection costs

Analytics accounts for the chat’s work: the agent’s requests and replies. Bulk collection costs — internet search, page downloads, text generation by models — are counted separately, in the harvesters’ statistics: each one shows how many tasks it handled and how much was spent. These are different line items: here is what the conversation with the agent costs, and in the [Harvesters](/agents/data/docs/harvesters)
 section — what collecting the data costs.

## Why you care

Automating data collection is not just a convenience — it is a cost item worth keeping under control. Analytics shows its payback: how much the agent “works” per day and what that costs — and where saving (a cheap model for routine scheduled tasks, for instance) lowers the bill without losing quality.

See also: [Chat with the Agent](/agents/data/docs/chat)
, [Harvesters](/agents/data/docs/harvesters)
.

[← Back to the documentation index](/agents/data/docs)