This HTML file is only the dashboard interface. The AgentTelemetry app needs to be running to collect data and manage local servers.
Open the running dashboard If the dashboard does not open, launch AgentTelemetry.app first, then try again.share of the selected range, and the tools that ran each
ranked by the selected measure · click one for its sessions
local hour × weekday, by tokens
what each tool's tokens were spent on
last 12 months · click a day to zoom in
tokens by month · click a month to zoom in
running total by tool · dashed = previous period
what each kind of token cost you
cache reads as a share of context tokens
cost per 1M tokens
stacked by tool
Account-wide limits · tools used today on this device
click a header to sort · a model used in several tools expands into its per-tool split
how often each model's edits land first time, and what a prompt costs on it · Claude Code & Codex
top 8 models per day
100% of each day
where each maker's models actually ran
dot = the tool that called it most
tool calls by category
each prompt and the work it set off
share of editing prompts with no edit → run → re-edit of the same file
which IDE or surface each tool ran in
calls, and tools used out of those offered
requests a Skill drove (Claude Code)
click a header to sort
by working directory
click a row to see its sessions
click a row or its title for detail
Counts follow the dashboard filters, date range and search. Only source-recorded markers identify subagents.
live files on disk right now
how expensively each tool stores its history
cumulative bytes by the day each log was last written
the files actually taking the space
AI data on disk that isn't turned into analytics