AI usage

AI usage, the last tab of Reports, shows where Limn’s own AI runs went for this workspace in the reporting window: reviews, story reviews, AI Fix, chat, diagrams and the rest. It also lists ways to use fewer tokens, worked out from your runs. The tab appears only while AI is on.

#What you see

• The totals: Tokens (input plus output, for Claude and Codex alike; tokens Claude read from its cache are not counted), Runs, and the cost. Runs that are not tied to a repository, such as This Sprint, Docs and summaries, count in every workspace, and a line says how many there are.
• Tokens over time: By day or By week, Claude and Codex drawn apart, with the number of runs.
• Where it goes: tokens by feature.
• Top pull requests and Top stories: where the most tokens went. A story counts its own story reviews plus every other run on its pull requests.
• Last 12 weeks: a weekly bar chart and its trend, whatever the reporting window is. The last bar is this week so far.
• Ways to use fewer tokens: suggestions with an estimate of what each would have saved in this window, and a link to the setting it changes, such as Auto review settings or Effort settings. Limn works these out from your runs; no AI writes them.

#How cost is shown

• Claude on a subscription (your Claude Code sign-in): the share of a week’s allowance, estimated from your plan. Limn never shows US dollars for a subscription.
• Claude on an API key: US dollars.
• Codex on a ChatGPT sign-in: tokens only.

#What you can do

1. Open Reports and choose AI usage.
2. Select a pull request or story in the top lists to open it.
3. Under Ways to use fewer tokens, press a suggestion’s link to open the setting it changes.

#Why it helps

AI review runs on your own Claude or Codex account. This tab shows you what Limn spends of it, per feature and per pull request, so you can turn down what is not worth it rather than guess.