Analytics
Overview
The Analytics page gives you a breakdown of your token usage and costs across models, agents, and tools. Every LLM API call is recorded locally in your browser's IndexedDB -- no data leaves your device.
Access Analytics from the chart icon in the home page header.
Summary Cards
At the top of the page, four cards show aggregated stats for the selected time range:
- Total Cost -- your total spending in USD, with a trend indicator comparing against the previous period
- Total Tokens -- combined input and output tokens, with a per-type breakdown
- API Calls -- total number of LLM requests
- Avg / Call -- average cost per API call
Time Range
Filter data by time range using the segmented control:
- 7d -- last 7 days
- 30d -- last 30 days (default)
- 90d -- last 90 days
- All -- all recorded data
The trend indicator on the Total Cost card compares the current period against the previous one (e.g., this 7 days vs. the prior 7 days).
Usage Over Time
A bar chart shows daily usage. Toggle between Cost and Tokens views:
- Cost -- daily spending in USD
- Tokens -- stacked bars showing input tokens (blue) and output tokens (green)
Hover over bars to see exact values in the tooltip.
Breakdowns
Per Agent
Shows which agents consume the most budget. Each agent displays a progress bar relative to the top spender, along with its total cost.
Per Model
Same layout as agent breakdown, but grouped by model ID (e.g., gpt-4.1, claude-sonnet-4-5-20250929). Useful for comparing cost efficiency across providers.
Tool Usage
Lists the top 10 most-called tools with their invocation count. This helps identify which MCP tools your agents rely on most.
Per-Chat Usage
Token counts and costs are also tracked per conversation. You can see usage inline in two places:
- Agent sidebar -- each chat shows its total tokens and cost below the title (e.g.,
1.2K tokens ยท $0.0012) - Chat list -- the same usage summary appears next to the timestamp, with a coin icon
Usage updates in real time after each message, so you always see the current cost of a conversation.
Session Tokens
While chatting, the chat window footer displays a running total of tokens used in the current session (since you opened the chat), along with an estimated cost based on the active model's pricing.
Clearing Data
Click the trash icon in the Analytics page header to delete all analytics records. A confirmation dialog appears before any data is removed. This action is irreversible.
How It Works
Every time an LLM call completes (including each round in the tool-calling loop), a usage record is saved to IndexedDB with:
- Input and output token counts
- Cost calculated from the provider's pricing
- Agent and model identifiers
- Names of any tools called in that round
- Chat ID (used for per-conversation aggregation)
All calculations happen client-side. No analytics data is sent to any server.