Model Usage

Zero-config CLI for analyzing AI model usage & costs across all local TUI tools — pi, Claude Code, Codex, Gemini CLI, OpenCode, Grok CLI, and more.

✨ Features

• 🪄 Zero Configuration - Auto-detects data from pi, Claude Code, Codex, Gemini CLI, OpenCode, Grok CLI
• 🔒 100% Local & Secure - No API keys required, reads local data only
• ⚡ Quick Analysis - View all usage with a single mu command
• 💰 Cost Tracking - Accurate cost calculation (pre-computed or LiteLLM pricing)
• 🎛️ Interactive Source Selection - TUI checkbox to pick which tools to query
• 🎯 Smart Project Detection - Auto-detects current project in project directories
• 📊 Multiple View Modes - Daily aggregated, per-message detailed, or by-date
• 🔍 Smart Filtering - Filter by time ranges, project names, and models
• 📈 Flexible Sorting - Sort by cost, time, tokens, or project name
• 🧩 Extensible - Add new TUI tool support in one file

🚀 Quick Start

# Install globally
npm install -g model-usage

# Run — auto-detects all your TUI tools automatically
mu

That's it. No configuration needed. On first run, you'll see an interactive checkbox to select which TUI tools to include, then your usage stats appear.

Other Installation Methods

# One-time usage
npx model-usage

# Local project install
npm install model-usage
npx model-usage

🛠️ Supported TUI Tools (auto-detected)

Tool	Data Source	Known Models
pi	~/.pi/agent/sessions/	deepseek-v4-pro, etc.
Claude Code	~/.claude/projects/	claude-sonnet-4-5, etc.
Codex CLI	~/.codex/sessions/	gpt-5, etc.
Gemini CLI	~/.gemini/tmp/<hash>/chats/	gemini-3-pro-preview, etc.
OpenCode	~/.local/share/opencode/opencode.db	deepseek-v4-pro, etc.
Grok CLI	~/.grok/grok.db	grok-4.3, etc.

All six are detected automatically — no config files needed.

📋 Usage

Source Selection

Control which TUI tools to query:

mu                        # Interactive checkbox to pick sources (in TTY)
mu --sources pi,opencode  # Only query pi and OpenCode
mu --sources all          # Query all available sources (skip prompt)

In non-interactive environments (piped, cron, etc.), all sources are auto-selected without prompting.

Basic Commands

mu                        # All usage for auto-detected project (interactive source picker in TTY)
mu -a, --all-projects     # All projects across all selected sources
mu -v, --version          # Version
mu -h, --help             # Help

mu -P, --projects         # List all projects
mu -M, --models           # List all models with LiteLLM pricing

🔍 Filtering

# Time filters
mu -t 7d                  # Last 7 days
mu -t 2h                  # Last 2 hours
mu -t 2024-07-01T14:30,2024-07-01T16:45  # Date range (ISO 8601)

# Project & model filters (partial matching supported)
mu -p myproject           # Projects containing "myproject"
mu -m sonnet              # Models containing "sonnet"
mu -m gpt-4               # Models containing "gpt-4"

# Combine
mu -t 7d -m sonnet -p myproject

📈 Sorting

mu -k cost -o desc        # Highest cost first
mu -k tokens -o desc      # Most tokens first
mu -k project -o asc      # Alphabetical by project
mu -k time -o desc        # Newest first

📊 View Modes

mu                        # Aggregated by project + date (default)
mu -d                     # Detailed: individual messages
mu --by-date              # By date only, all projects combined

⭐ Config

mu --disable-github-prompt  # Hide GitHub star prompt permanently
mu --enable-github-prompt   # Show it again

📊 Sample Output

🔍 Options applied:
  Project: my-website (auto-detected)
  Sort: cost ↓

┌──────────┬─────────────┬──────────┬────────┬────────┬──────────────┬────────────┬─────────────────┬──────────┬────────────┐
│ Time     │ Project     │ Messages │ Input  │ Output │ Cache Create │ Cache Read │ Model           │ Total    │ Cost (USD) │
├──────────┼─────────────┼──────────┼────────┼────────┼──────────────┼────────────┼─────────────────┼──────────┼────────────┤
│ 7/1/2026 │ my-website  │       15 │  1,200 │    400 │            0 │     20,000 │ gpt-4.1         │   21,600 │      $0.09 │
├──────────┼─────────────┼──────────┼────────┼────────┼──────────────┼────────────┼─────────────────┼──────────┼────────────┤
│ 6/30/2026│ my-website  │       20 │    800 │    300 │            0 │     15,000 │ gpt-4.1         │   16,100 │      $0.06 │
├──────────┼─────────────┼──────────┼────────┼────────┼──────────────┼────────────┼─────────────────┼──────────┼────────────┤
│ TOTAL    │             │       35 │  2,000 │    700 │            0 │     35,000 │                 │   37,700 │      $0.15 │
└──────────┴─────────────┴──────────┴────────┴────────┴──────────────┴────────────┴─────────────────┴──────────┴────────────┘

🛠️ Requirements

• Node.js >= 22.0.0 (uses built-in node:sqlite)
• At least one supported TUI tool with conversation history

🔒 Privacy & Security

• 100% Local — reads only from your local disk
• Minimal Network — only fetches model pricing from LiteLLM (cached 1h)
• No API Keys — no authentication required
• Privacy First — your data never leaves your machine

🧩 Extending

To add support for a new TUI tool, create src/sources/mytool.ts:

export const name = 'mytool';

export function isAvailable(): boolean { /* sync check */ }

export async function readSessions(): Promise<UsageResult> {
  // return { messages, totals }
}

export async function getProjects(): Promise<ProjectsResult> {
  // return { projects, messageCount }
}

No other files need changes — auto-discovery picks it up. See src/sources/common.ts for shared utilities.

📄 License

MIT

🔗 Links

• NPM
• GitHub
• Issues