🚨 Active Supply Chain Attack:node-ipc Package Compromised.Learn More
Socket
Book a DemoSign in
Socket

@jeganwrites/claudash

Package Overview
Dependencies
Maintainers
1
Versions
27
Alerts
File Explorer

Advanced tools

Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

@jeganwrites/claudash

Claude Code usage intelligence dashboard

latest
Source
npmnpm
Version
3.3.1
Version published
Maintainers
1
Created
Source

Claudash

Claude Code usage intelligence dashboard. Reads your local session files, detects waste patterns, generates fixes, and measures whether they worked.

Dashboard screenshot

What it does

Most Claude Code usage tools tell you how many tokens you spent. Claudash tells you why and what to do about it.

It detects four waste patterns in your sessions — repeated file reads, stuck retry loops, sessions that ran too long without compacting, and cost outliers — then generates targeted CLAUDE.md rules to fix them and measures the before/after difference.

Prerequisites

Before installing Claudash, you need:

RequirementWhyHow to check
Claude CodeClaudash reads its session filesclaude --version
Python 3.8+Claudash is written in Pythonpython3 --version
Node.js 16+Required for npx install method onlynode --version

Claude Code must have run at least one session before Claudash will show any data. Sessions are stored in:

  • macOS / Linux: ~/.claude/projects/
  • Windows: %APPDATA%\Claude\projects\
  • WSL2: /mnt/c/Users/<username>/AppData/Roaming/Claude/projects/

Install

Choose the method that works for your setup:

Method 1 — npx (fastest, no install needed)

npx @jeganwrites/claudash

Requires Node.js 16+. Downloads and runs without permanent installation.

Method 2 — npm global install

npm install -g @jeganwrites/claudash
claudash dashboard

Installs permanently. Run claudash from anywhere.

Method 3 — Homebrew (macOS / Linux)

brew tap pnjegan/claudash
brew install claudash
claudash dashboard

No Node.js required. Python only.

Method 4 — Git clone (full control)

git clone https://github.com/pnjegan/claudash.git
cd claudash
python3 cli.py dashboard

No npm, no Node.js required. Best for development or customisation.

Windows users

Claudash runs on Windows via WSL2 (Windows Subsystem for Linux). Native Windows support is in progress.

Setup WSL2:

# In PowerShell (as administrator)
wsl --install

# Restart, then open Ubuntu from Start menu
# Install Node.js inside WSL
curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt-get install -y nodejs

# Install Claudash
npx @jeganwrites/claudash

Claudash will automatically detect your Windows Claude Code sessions at /mnt/c/Users/<username>/AppData/Roaming/Claude/projects/.

macOS users

Quickest path:

# Install Node.js if you don't have it
brew install node

# Run Claudash
npx @jeganwrites/claudash

Or use Homebrew install (no Node.js needed):

brew tap pnjegan/claudash
brew install claudash
claudash dashboard

Linux users

# Ubuntu / Debian — install Node.js
curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt-get install -y nodejs

# Run
npx @jeganwrites/claudash

First run

# 1. Start the dashboard (auto-opens browser)
claudash dashboard

# 2. If no data appears, run the scanner manually
claudash scan

# 3. Check what's in your session files
claudash stats

Dashboard opens at http://localhost:8080

If the port is in use:

PORT=9090 claudash dashboard

What you'll see

After your first scan:

  • Summary bar — total sessions, total cost, cache hit rate
  • Project breakdown — which projects cost the most
  • Efficiency score — 0-100 grade across 5 dimensions
  • Active insights — specific problems with fix suggestions
  • Waste events — detected patterns with token cost
  • Fix tracker — before/after measurement for applied fixes

A score below 50 is not unusual. The score measures how efficiently you're using your context window, not whether the work was good.

Backup and recovery

# Manual backup (creates snapshot + JSON export)
claudash backup

# Restore from backup
claudash restore --file ~/.claudash/backups/claudash-20260418.db

# Automated hourly backup (add to crontab)
0 * * * * cd /path/to/claudash && python3 cli.py backup --quiet

Your irreplaceable data is in the fixes and fix_measurements tables. Everything else regenerates from your session files.

Privacy and data handling

Claudash reads your Claude Code session files stored locally on your machine.

  • Nothing is uploaded — all data stays in data/usage.db on your machine
  • No telemetry — no usage tracking, no analytics, no external calls
  • No accounts — no login, no email, no cloud sync
  • Session files contain your full conversation history — Claudash reads them to extract token counts and tool calls only, not conversation content

The database is created with restricted file permissions (0600 on Unix systems). API keys for fix generation are stored locally in the same database.

For team or cloud deployment, see SECURITY.md.

Troubleshooting

Dashboard shows no data

  • Run claudash scan first
  • Check that ~/.claude/projects/ contains .jsonl files
  • On Windows, run inside WSL2

Port 8080 already in use

PORT=9090 claudash dashboard

Python not found

# macOS
brew install python@3.11

# Ubuntu
sudo apt install python3

Node.js version too old Claudash requires Node.js 16+. Check with node --version. Update via brew upgrade node (macOS) or via nvm.

Permission denied on database

chmod 600 data/usage.db

WSL2 can't find Windows sessions Claudash looks for sessions in /mnt/c/Users/<username>/AppData/Roaming/Claude/projects/. Confirm that path exists: ls /mnt/c/Users/ and find your username.

Contributing

Pull requests welcome. Especially:

  • Windows native support (without WSL2)
  • Firefox session key support for browser tracking
  • Additional waste pattern detectors
git clone https://github.com/pnjegan/claudash.git
cd claudash
python3 cli.py dashboard  # no install needed

License

MIT — use it, fork it, build on it.

All data stays on your machine. Zero pip dependencies. One command install.

Keywords

claude

FAQs

Package last updated on 18 Apr 2026

Did you know?

Socket

Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.

Install

Related posts