Understand Work
Analytics
Use local session history to understand time, token use, concrete output, tool activity, and risk signals for the selected project.
Open analytics
- Select a project.
- Choose the Analytics badge in the project header.
- Use the summary cards and charts to compare sessions and engines.
Only sessions associated with the selected project are included. Claude Code is marked CC; Codex is marked CX.
What the metrics mean
| Metric | How to read it |
|---|---|
| Tokens | Approximate model input and output recorded by the CLI. |
| Active time | Estimated working time from session events; long idle gaps are excluded. |
| Artifacts | Files, tasks, or other concrete outputs detected in a session. |
| LOC | Lines added and removed where change data is available. |
| Tool calls | Recorded tool activity, including failures and permission checkpoints. |
| Risks | Signals worth reviewing, not automatic proof that a session was unsafe. |
Useful views
- Activity heatmap shows when work happened.
- Tokens vs Output helps identify expensive exploration or stalled work.
- Agent Comparison compares Claude Code and Codex on the selected project.
- Session Productivity connects individual sessions to concrete output.
- Tool Usage surfaces volume, failures, commands, and risk-heavy activity.
Interpret with context
A high token count can be appropriate for research or a difficult migration. A permission-heavy session may simply be installing dependencies or working across repositories. Use analytics to find sessions to review, then inspect their conversation and Git diff before drawing conclusions.
Partial data is normal: What appears depends on what each CLI recorded. A missing metric does not necessarily mean no work happened.