Usage per subscription, cost per merged issue, lines per person, people next to agents. Every agent commit credited to whoever started it. And when a number is unknown, it says unknown.
Overview, usage, efficiency, development, quality and person. Each answers one thing a lead actually asks on a Monday. Click through.
The daemon reads quota and usage from Claude Code, Codex and Grok on the machine they run on. Insight shows it per subscription, so you see who is about to hit a wall before the agent does.
Commits, lines and PRs per person, with the agent's share right next to it. The GitHub App attributes every agent commit to the person who started the run, so nobody's work disappears into a bot account.
How commits are linked: the GitHub App →
Save mode is on by default. A run starts on a cheaper profile, and if it fails, the retry escalates to a stronger one. You pay for the big model only on the work that needed it.
When a number isn't known, Insight says unknown. It never fills the gap with a zero that drags your averages down and makes a chart lie.
Bring your history from Linear: migrate →
Pair a machine and Insight starts counting. Free while we're in beta.