Dashboard
The dashboard summarizes how you use Claude Code across all uploaded sessions.
Choose what to look at#
At the top, pick a time range (7, 30, 90, 365 days or all time) and a project or all projects. Everything below follows these two choices. Click a project in the table, or a program in the shell list, to filter by it.
If a range is empty, pick a longer one or import your history.
What you see#
| Section | Tells you |
|---|---|
| Totals | Sessions, prompts, tool calls, failure rate, output and cache tokens, peak context, alerts. |
| Activity | Per day. Switch the measure: calls, sessions, tokens, failures or alerts. |
| When you work with Claude | A weekday by hour heatmap in your local time. |
| Tool categories | Shell, read, write, search, web, agent… and their share of calls. |
| Tools | Per tool: calls, failures, average and p95 duration, total time. p95 means 95% of calls were faster than this. |
| Projects | Per folder: sessions, calls, failures, output tokens, alerts, last activity. |
| Top shell programs | The programs the model ran most. Click one to search it. |
| Alerts by rule | Which rules matched how often. |
| Peak context per session | How large the context grew in each session. |
| Models | Sessions that used each model. A session can use several, for example for subagents. |
| Biggest sessions | Sort by tokens, calls, duration or context. Click to open. |
Tips#
- A rising failure rate or tool with many failures points to something to fix in your setup or instructions.
- Sessions with a very high peak context are slow and costly. Starting a fresh session sooner can help.
- Use Alerts by rule to see which risky action shows up most.