By employee
Usage is attributed to a person, not only to a shared API key. A manager can see Maya’s Cursor prompts separately from Andre’s Codex prompts.
Employee AI usage metrics
See every employee’s AI usage, for every coding agent.
AI Usage Metrics records one row each time an employee finishes a prompt. The row names the person, the agent, and the input, output, and cached tokens. Company leaders open one report instead of a separate console for each tool.
View the demo with your company email
The demo is sample data. A company email is required. Personal inboxes are declined.
By agent
Cursor, Claude Code, Codex, and GitHub Copilot keep their own names. The token fields stay the same, so totals can be compared.
By prompt
Each completed reply is one usage row. Turns are the number of rows. A skipped prompt is absent, so the count is recorded completions.
What a usage row contains
| Field | Meaning |
|---|---|
| Employee | The person who ran the agent. |
| Agent | Cursor, Claude Code, Codex, Copilot, or another coding agent. |
| Input tokens | Context read for that prompt. |
| Output tokens | The reply itself. |
| Cached tokens | Context served from cache instead of reread in full. |
Prompt text stays on the employee’s machine. The report stores the counts.