TraceAI CLI

Track where your AI spend is going.

One-command CLI to trace prompt usage, token spend, and model waste across OpenAI, Claude, and Gemini.

Installation

# Run directly with npx (no install required)
npx traceai-cli

# Or install globally
npm install -g traceai-cli

Quick Start

# Set your API keys
export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="sk-ant-..."
export GOOGLE_API_KEY="AIza..."

# Run TraceAI
traceai

Usage

# Basic scan (last 30 days)
traceai

# Generate full report
traceai --report

# Scan specific time period
traceai --days 7

# Output as JSON
traceai --json

# Output as CSV
traceai --report --csv

# Scan specific provider only
traceai --provider openai

# Configure API keys interactively
traceai --configure

Commands & Options

| Option | Description | |--------|-------------| | --report, -r | Generate full report and save to file | | --json | Output as JSON format | | --csv | Output as CSV format | | --days <n> | Number of days to analyze (default: 30) | | --provider <name> | Scan specific provider (openai, anthropic, gemini) | | --output <path> | Custom output path for report | | --configure | Configure API keys interactively | | --help | Show help | | --version | Show version |

Configuration

Environment Variables (Recommended)

export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="sk-ant-..."
export GOOGLE_API_KEY="AIza..."

Config File

Create ~/.traceai/config.json:

{
  "providers": {
    "openai": {
      "api_key": "sk-...",
      "org_id": "org-..."
    },
    "anthropic": {
      "api_key": "sk-ant-..."
    },
    "gemini": {
      "api_key": "AIza..."
    }
  }
}

Sample Output

╭──────────────────────────────────────────╮
│                                          │
│   TraceAI v1.0.0                         │
│   Track where your AI spend is going     │
│                                          │
╰──────────────────────────────────────────╯

✓ OpenAI API connected
✓ Anthropic API connected
! Gemini API not configured

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
USAGE SUMMARY (Last 30 days)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Provider       Model              Tokens       Cost
─────────────────────────────────────────────────
OpenAI         gpt-4-turbo       1,247,832    $48.72
OpenAI         gpt-3.5-turbo       892,104     $1.78
Anthropic      claude-3-opus       234,521    $14.07
─────────────────────────────────────────────────
TOTAL                            2,374,457    $64.57

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
WASTE DETECTED
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

! 23 retry loops detected            ~$8.40 wasted
! 47 duplicate prompts found         ~$3.20 wasted
! 12 high-cost/low-output queries    ~$6.80 wasted
─────────────────────────────────────────────────
Potential savings: $18.40/month (28.5%)

Features

• Multi-Provider Support - OpenAI, Anthropic Claude, Google Gemini
• Waste Detection - Identify retry loops, duplicate prompts, expensive queries
• Cost Analysis - Detailed breakdown by model and provider
• Privacy-First - All data stays local, no telemetry by default
• Multiple Formats - Markdown, JSON, CSV output

Privacy & Security

• No outbound telemetry by default
• API keys are never transmitted to external servers
• All analysis runs locally on your machine
• Optional anonymous analytics (opt-in only)

Requirements

• Node.js >= 16.0.0
• API keys with billing/usage access

License

MIT License - see LICENSE for details.

Links

• Website: slab.twc.sh/traceai
• Documentation: slab.twc.sh/traceai/docs.html
• Issues: GitHub Issues

Part of the Alt.QA reliability suite by TAO.ai