claudeactivity

Activity visualizer for Claude Code. Scans your local Claude session logs and generates a clean HTML report showing which repos, files, and tools you use most.

Usage

No install needed:

npx claudeactivity scan

Or install globally:

npm install -g claudeactivity
claudeactivity scan

The report opens in your browser automatically. If it doesn't, the path is printed — copy it and open manually.

The report is a tabbed dashboard with a date-range filter (7d / 30d / 90d / All) that recomputes every section live.

What it shows

Overview

• Activity timeline — daily file accesses across the selected range
• Tool breakdown — Read / Edit / Search / Exec proportions
• Coding persona — inferred from your tool usage ratios

Projects

• Project activity — all repos ranked by access count, sessions, and avg session depth
• Repository breakdown — per-repo cards with session stats and top files

Files

• Most accessed files and most edited files (edits/writes only)
• Directory hotspots — most accessed directories
• Language breakdown — share of activity by file extension

Sessions

• Time-of-day heatmap — when you use Claude, by hour and day of week
• Session duration — average, median, longest, and a length histogram

Tokens (read directly from session logs — no estimates)

• Total tokens and cache hit rate
• Tokens by model (Opus / Sonnet / Haiku) and by project
• Token composition — fresh input / output / cache write / cache read
• Token usage over time

Options

claudeactivity scan                        # scan all projects, open report
claudeactivity scan --no-open              # generate but don't open browser
claudeactivity scan --project              # scope to current directory only
claudeactivity scan --output /tmp/reports  # custom output directory
claudeactivity scan /custom/claude/path    # custom Claude projects directory

Reports are saved to:

.claudeactivity/latest-report.html                     # always overwritten
.claudeactivity/reports/report-YYYY-MM-DD-HH-MM.html  # timestamped copy

How it works

Reads JSONL session logs from ~/.claude/projects/ (or ~/.config/claude/projects/), extracts file accesses, tool usage, and token counts from each session, and renders a self-contained HTML report with no external dependencies. All filtering and charts run client-side, so the date filter works without regenerating the report.

Requirements

• Node.js 18+
• Claude Code installed and used at least once

Platform support

| Platform | Status | |---|---| | macOS | ✓ | | Linux | ✓ | | WSL | ✓ | | Windows | ✓ |