Insights

Insights rolls up every run on the server. Pick a range (7, 30 or 90 days) at the top.

Headline numbers

Tile	What it counts
Reviews	Agent runs started in the range.
Tokens	Real tokens those runs used, parsed from the Claude Code stream log. Cached re-runs count zero.
Kept as-is	Share of findings posted unchanged out of all findings that reached a post decision.
PRs · reviewers	Distinct PRs and distinct people who have run reviews, all-time.

Panels

• Reviews per day — a small bar chart for the range.
• What happened to findings — kept / edited / dropped for the range. This is the learnings loop’s raw material.
• Findings by severity — blocker, should-fix, nit, question, all-time.
• By reviewer — runs per login, all-time.
• By model — runs per model, with tokens.
• Agreement across reviewers — of the PRs where more than one person ran an independent review, how often the same finding was raised. Independence-weighted: it counts only when the runs used a different skill, model or effort. A high number is a signal about the findings; a low one is a signal about the skills.
• Cycle time — how long between a review being requested and the comments being posted, for PRs that reached a post. Lagging by nature.

What it is for

Deciding where to spend tokens (which effort and model actually change the keep rate), and which skill deserves to become the team default. It is not a performance review of people; the per-reviewer panel exists so you can see who is carrying the queue, not to rank them.