Repository navigation
Usage: activity analytics for agent time, tool calls, shell commands, model behavior, languages and plan value #39301
mfroembgen
started this conversation in
Ideas
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Problem
The Usage page shows how much was spent: tokens, cost at API prices, tokens by type, cache savings, and daily and hourly charts per provider and model, next to plan limits. It doesn't show how the agents work. I use a separate menu-bar tool that reads the same Claude and Codex session files to answer questions like these:
T3 already has every turn, tool call, file change and model choice, so it could answer these itself, across all its providers.
Data points missing in T3 Code today
All of them should work over a selectable range (7 days, 30 days, all time) and split by provider and model.
Activity overview
Shell commands
These record program names only, never arguments or paths.
head,cat,sed,git,rg,grep,pythonand so on), split by provider.headortailgrepinstead ofrgpython -c)git statuscalls per commitModel behavior
Languages
Plan value
Sharing
Related
Suggested shape
Add an Activity or Insights tab next to Cost and Tokens on the Usage page, using the same filters for provider, model, environment and period. A model shows up in a chart only after it has enough samples in the range, for example 10 prompts or 30 commands, so a few runs don't skew the comparison.
All reactions