Analyze Claude Code usage and tool trends

Analyze usage patterns and trends from Agent Monitor analytics

Weekly review of what's actually happening in your system — peak usage patterns, which tools are most relied on, error spikes.

• · Time period: 'last 7 days', 'last 30 days', 'last quarter'
• · Optional: 'peak hours', 'tool trends', 'model usage' for specific analysis

• · Daily activity trend (sessions/day, events/day, week-over-week delta)
• · Token volume trends (input, output, cache_read, cache_write over time)
• · Tool usage ranking (top 20 tools with bar chart data)
• · Model distribution (by frequency and subagent type)
• · Session health distribution (completion, error, abandoned rates)
• · Event type distribution (PreToolUse, PostToolUse, errors, compactions)

• · Agent Monitor API: /api/analytics, /api/stats, /api/sessions

• · Agent Monitor running with analytics data
• · Daily trends require at least 7 days of data
• · Tool usage requires PreToolUse events to be logged

• · Peak hours analysis only meaningful if system has clear usage patterns (e.g., not 24/7 background)
• · Tool rankings can be misleading if tool names are not consistent (Bash vs bash, ReadFile vs Read)
• · Model distribution doesn't account for cost (frequency ≠ spending)
• · Event type distribution misses agent-internal tool calls (only sees public PreToolUse/PostToolUse)

• · 365-day historical data for daily_events and daily_sessions endpoints
• · Week-over-week delta calculation to detect trend direction
• · Tool diversity metric (unique tools used)
• · Subagent count from total_subagents in analytics
• · Event type distribution from event_types breakdown (PreToolUse, PostToolUse, APIError, Compaction, etc.)
• · ASCII trend indicators (▲▼→) for visual quick-read
• · Period comparison with markdown tables