See what your AI coding agents store on your machine.
agentlens scans ~/.claude/ for data stored by Claude Code and shows you what's there: memory files, token costs, feature flags, session transcripts, and your total data footprint.
npx agentlens scanNo API keys, no accounts, no config. Reads files already on your disk.
What Claude remembers about you across projects: user profiles, feedback, project context. Scans for sensitive data (API keys, secrets) accidentally stored in memory files.
Token usage and spending broken down by model and project. Shows which models consumed tokens and your cache hit rate.
Active feature flags cached on your machine via GrowthBook. Shows what's toggled on your account: the flag names and values Anthropic has set.
Analyzes your conversation transcripts: message counts, tool usage frequency, which models were used, and total transcript size.
Full data footprint audit: how many files Claude Code stores, how much space they take, and what categories they fall into (transcripts, memory, config).
agentlens scan # Full scan (default)
agentlens memory # What Claude remembers about you
agentlens costs # Token usage and spending
agentlens features # Active feature flags and config
agentlens sessions # Session history and tool usage
agentlens privacy # Data footprint auditagentlens clean # Delete memory files (--dry-run, --project, --older-than)
agentlens redact # Find secrets in memory files (--fix to redact them)
agentlens diff save # Save a snapshot of current state
agentlens diff show # Compare the two most recent snapshots
agentlens export # Export all data to portable JSON (--memory, --sessions, --costs)AI coding agents store a lot of data on your machine: conversation transcripts, persistent memory files, feature flags, cost data. Most of it is invisible unless you go digging through ~/.claude/ yourself.
agentlens makes it easy to see what's there.
MIT