About this project

The Claude Code Dashboard for Personal Analytics provides a private, offline view of how you interact with Claude Code. It reads the JSONL session transcripts stored in ~/.claude/projects/ (never modifying them), extracts metrics such as token usage per model, API‑equivalent cost, session duration, tool invocations, skill usage, and git‑related details, and caches the results in a local SQLite database for fast incremental updates. A single‑page web interface served by a lightweight FastAPI server presents overview KPIs, daily cost charts, token share by model, top tools and skills, project‑level totals with tagging and hiding controls, a searchable session list with detail views, an activity heatmap (weekday × hour), and a calendar‑style week view that color‑codes sessions by project tag. Configuration is handled via environment variables or command‑line flags for host, port, authentication token, database location, and optional redaction of home paths in exported data. The tool can be run directly from PyPI via uvx or pipx, installed persistently with uv tool or pipx, or launched from a source checkout using uv run python -m claude_dashboard. Optional static export allows hosting a self‑contained version without live access to logs. All processing stays on the user's machine unless they choose to expose the server on a trusted network or deploy the static export elsewhere.