Agent Stats
Agent Stats gives you per-agent analytics cards showing command counts, success rates, average durations, risk breakdowns, and the most-used commands — so you can understand how each AI agent is using your terminal.
Usage
suv agent statsThis displays an interactive TUI with analytics cards for each detected agent. The data comes from commands recorded by Suvadu's agent tracking.
Flags
| Flag | Description |
|---|---|
-d / --days <N> | Number of days to include in the stats (default: 30) |
--executor <NAME> | Show stats for a specific agent only (e.g., claude-code, cursor) |
--text | Output plain text instead of the interactive TUI (useful for piping or scripting) |
What the Stats Show
Each agent gets its own analytics card displaying the following metrics:
Summary Metrics
- Total commands — the number of commands this agent has executed in the selected time period
- Success rate — percentage of commands that exited with code 0
- Average duration — mean execution time across all commands from this agent
Risk Breakdown
A breakdown of commands by risk level, showing how many commands fell into each category:
- Critical — destructive operations like
rm -rf /orDROP TABLE - High — potentially dangerous operations like
chmod 777ornpm install - Medium — operations that modify state like
git resetordocker run - Low — minor modifications like file writes or branch operations
- Safe — read-only operations like
git status,ls, orgrep
Top Commands
A ranked list of the most frequently used commands by this agent, showing the command text and how many times it was run.
Top Directories
A ranked list of the directories where this agent executed the most commands, showing which projects or areas of your filesystem the agent works in most.
High-Risk Commands Table
A table listing commands from this agent that were classified as High or Critical risk, with timestamps and exit codes. In 0.4.0 the table uses the available width instead of truncating command text at 30 characters. This gives you a quick audit trail of the most consequential actions each agent has taken.
Keyboard Shortcuts
Press 1 for Today, 2 for 7 days, 3 for 30 days, or 4 for All time. Use Up/Down to scroll and q or Esc to quit. Agent cards now use rounded borders consistent with the other TUI screens.
Examples
Last 7 days of agent stats
suv agent stats --days 7Shows analytics cards for all agents over the past week. Useful for a weekly review of agent activity.
Cursor-specific stats as plain text
suv agent stats --executor cursor --textOutputs Cursor's analytics in plain text format, suitable for piping to other tools or including in reports.
Full 30-day overview
suv agent statsDefault view: 30-day analytics for all detected agents in the interactive TUI.
Frequently Asked Questions
How is the success rate calculated?
Success rate is the percentage of an agent's commands that exited with code 0. Commands that exited with any non-zero status count against it, so a lower success rate points to an agent hitting more failures during its runs.
Which agents does suv agent stats track?
It shows a card for every agent Suvadu has recorded commands from via agent tracking — for example claude-code and cursor. Use --executor <NAME> to narrow the view to a single agent.
How do I change the time range?
Pass --days <N>. The default is the last 30 days, so suv agent stats --days 7 gives you a weekly review and a larger number widens the window.
Can I use agent stats in a script?
Yes. Add --text to print plain-text output instead of the interactive TUI, which you can pipe into other tools or drop into a report.
How is this different from the Agent Dashboard and Agent Reports?
Agent Stats is aggregate analytics over a time window. The Agent Dashboard is for real-time monitoring of live activity, while Agent Reports produces exportable reports that include session replay.