User Guide

Usage

Track costs, tokens, and API call analytics across chat and agents

Usage

The Usage page gives you full visibility into how much your Claude interactions are costing and where those costs are coming from. It tracks every API call across chat conversations, agent runs, and memory extraction.

Time Period

Use the 24h, 7d, or 30d buttons in the navbar to change the reporting window. Your selected period is persisted between sessions.

Stats Cards

Five summary cards appear at the top of the page:

CardWhat It Shows
Total CostThe cumulative USD cost for the selected period.
Input TokensTotal tokens sent to Claude (your messages, system prompts, context).
Output TokensTotal tokens generated by Claude (responses, tool calls).
API CallsThe number of individual API requests made.
Avg / CallThe average cost per API call.

Token counts are formatted for readability -- for example, 1,500,000 tokens displays as "1.5M" and 45,000 as "45.0K."

Cost Over Time Chart

A line or bar chart showing your spending over the selected period. You can toggle between two granularity levels:

  • Daily -- One data point per day.
  • Hourly -- One data point per hour (most useful with the 24h period).

The chart adjusts automatically based on your timezone.

Cost by Source

A donut chart breaking down spending by source category:

SourceWhat It Includes
ChatConversations via the Chat page or Assistant Panel.
AgentScheduled agent runs.
MemoryMemory extraction processing.

This helps you understand whether most of your costs are coming from interactive use or automated agents.

Top Consumers

A ranked list of the individual items (conversations or agents) that have consumed the most tokens or cost. This is useful for identifying particularly expensive agents or long-running conversations.

Usage Records Table

Below the charts, a detailed table lists every individual API call with:

ColumnDescription
SourceColor-coded badge (Chat, Agent, or Memory).
NameThe conversation title or agent name associated with the call.
InputInput token count.
OutputOutput token count.
CostUSD cost for this specific call.
DurationHow long the API call took.
TimeWhen the call was made, shown as relative time.

Filtering Records

The records table has its own filter controls:

  • Source filter -- Show only Chat, Agents, or Memory records.
  • Search -- Filter by name (conversation title or agent name). Results update as you type.

Pagination

Records are paginated with 20 items per page. Navigation controls appear at the bottom when there are multiple pages.

Check this page regularly to spot unexpected cost spikes. If an agent is running more often or more expensively than expected, you can adjust its budget limit or schedule on the Agents page.

See API Reference for endpoint details.