ゆる草

How does Claude Code recover from errors? 99% self-recover rate. Track retry, investigate, fix, rollback, and pivot patterns across all your sessions.

Forecast your Claude Code usage — projected month-end hours, ghost days, and streak

Compare two time periods of Claude Code activity — week over week, month over month

ASCII receipt of your AI's daily work from Claude Code proof-log. The AI never clocks out.

AI activity report for your Claude Code Ghost Days. Shows what your AI did while you were offline.

Your Claude Code AI Productivity Score — a single 0-100 number you can share

AI-generated daily standup from Claude Code proof-log files

Time spent per project in Claude Code — ranked by hours, you vs AI

GitHub-style AI activity heatmap from Claude Code proof-log files

Show files changed by AI that need human review — reads activity-log.jsonl from Claude Code

Watch your active Claude Code session in real-time. Token usage, cache efficiency, burn rate, and cost estimate. Zero dependencies.

Project your Claude Code API cost to month-end

Generate a weekly AI activity report from Claude Code proof-log files.

Generate a Claude Code stats badge for your GitHub README — streak, hours, autonomy ratio

CLI diagnostic for your Claude Code setup. 20 checks across 6 dimensions.

Streak risk notifier for Claude Code — cron-friendly warning before your streak dies

How long can Claude Code go without an error? Median 12 successful tool calls between errors. Longest streak: 829. Bash breaks 52% of streaks.

Discover your Claude Code developer archetype. Diagnoses your coding style from real usage patterns.

One command to make Claude Code safe for autonomous operation. 8 hooks: destructive blocker, branch guard, force-push protection, secret leak prevention, syntax checks, and more.

Audit your Claude Code skills: token overhead, usage frequency, prune candidates

See what your AI did while you were gone. Shows git commits from Ghost Days — days AI ran Claude Code autonomously while you were offline.

How fast does Claude Code actually work? Tool execution rate (tools/hour), pace tiers, and burst session analysis.

Does Claude Code warm up or fade? Tool execution rate progression within sessions — early, mid, and late phase analysis.

What's the rhythm of a Claude Code session? Intra-session timing gap distribution — instant bursts, quick cycles, deliberate pauses.

Analyze which Claude Code tools appear together in the same session — co-occurrence and affinity matrix

Which files does Claude Code touch the most? Ranks files by Edit/Write/Read frequency across all sessions.

How does Claude's context window grow within a session? Context window usage analysis — size tiers, growth patterns, and compaction pressure.

See how full your Claude Code context window is — reads token usage from session transcripts

Analyze Claude Code tool burst patterns — how many consecutive calls does each tool make?