ai-economic-tracker（AI 经济追踪器）

AI Economic Tracker

Track daily costs, income, and net worth for AI agents. Implements economic pressure-driven decision making: "work or learn" based on financial runway.

Inspired by HKUDS/ClawWork research on economic incentives for AI agents.

Features

• Balance Tracking: Monitor current balance, total income, and total costs
• Survival Status: Automatic status classification (thriving/stable/struggling/critical/bankrupt)
• Runway Calculation: Days remaining before funds run out
• Service Valuation: Estimate task value using BLS wage data
• Work/Learn Decision: Economic pressure-driven task prioritization
• Daily Reports: Formatted economic status reports
• JSONL Logs: Append-only transaction history

Use Cases

1. Agent Cost Management: Track API costs, compute resources, and operational expenses
2. Revenue Tracking: Log income from completed tasks or services
3. Economic Decision Making: Decide whether to work (earn) or learn (invest) based on financial status
4. Service Pricing: Estimate fair pricing using US Bureau of Labor Statistics wage data
5. Financial Monitoring: Daily/weekly economic health checks

Installation

clawhub install ai-economic-tracker

Usage

Command Line

# View current status
python3 ~/.openclaw/workspace/skills/ai-economic-tracker/tracker.py status

# Daily report
python3 ~/.openclaw/workspace/skills/ai-economic-tracker/tracker.py report

# Initialize balance
python3 ~/.openclaw/workspace/skills/ai-economic-tracker/tracker.py init 1000.0

# Record income
python3 ~/.openclaw/workspace/skills/ai-economic-tracker/tracker.py income 50.0 "task_completion" "Completed data analysis"

# Record cost
python3 ~/.openclaw/workspace/skills/ai-economic-tracker/tracker.py cost 15.0 "api_usage" "OpenAI API calls"

# Estimate service value
python3 ~/.openclaw/workspace/skills/ai-economic-tracker/tracker.py estimate software_developer 2.5

# Get work/learn decision
python3 ~/.openclaw/workspace/skills/ai-economic-tracker/tracker.py decide

From OpenClaw Agent

# In your agent workflow
exec("python3 ~/.openclaw/workspace/skills/ai-economic-tracker/tracker.py report")

Cron Integration

Add to your OpenClaw cron for daily monitoring:

openclaw cron add "0 9 * * *" "python3 ~/.openclaw/workspace/skills/ai-economic-tracker/tracker.py report" --label "daily-economic-report"

Configuration

Set environment variables to customize:

# Data directory (default: ~/.openclaw/workspace/data/economics)
export ECONOMIC_TRACKER_DATA_DIR="/custom/path/to/data"

# Daily costs (default values shown)
export ECONOMIC_TRACKER_ELECTRICITY_DAILY=0.50
export ECONOMIC_TRACKER_INTERNET_DAILY=1.50

# Survival thresholds (default values shown)
export ECONOMIC_TRACKER_THRIVING=5000
export ECONOMIC_TRACKER_STABLE=1500
export ECONOMIC_TRACKER_STRUGGLING=500
export ECONOMIC_TRACKER_CRITICAL=0

Data Storage

All data stored in JSONL format (append-only):

• balance.jsonl: Balance snapshots with timestamps
• daily_log.jsonl: Cost transactions
• income_log.jsonl: Income transactions

Default location: ~/.openclaw/workspace/data/economics/

Status Levels

BLS Wage Reference

Built-in hourly wage data for service valuation:

• Computer Systems Manager: $90.38/hr
• Financial Manager: $86.76/hr
• Software Developer: $69.50/hr
• Financial Analyst: $47.16/hr
• Market Research: $38.71/hr
• Data Analyst: $52.00/hr
• General Operations: $64.00/hr
• Customer Service: $22.00/hr

Work/Learn Decision Logic

The tracker implements economic pressure-driven decision making:

• Critical/Struggling (< $1,500): Must work to earn money
• Stable ($1,500 - $5,000): 70% work, 30% learn
• Thriving (> $5,000): 50% work, 50% learn

Use decide command to get recommendation based on current balance.

Example Output

📊 经济日报 | 2026-02-26
========================================
💰 余额: $1,234.56
📈 总收入: $2,500.00
📉 总支出: $1,265.44
💵 净利润: $1,234.56
🔥 日消耗: $2.00
⏳ 跑道: 617 天
🔵 状态: STABLE
========================================

Dependencies

Zero external dependencies. Uses only Python standard library:

• json
• os
• datetime
• pathlib

Security

• No API keys required
• All data stored locally
• No network requests
• Configurable via environment variables (no hardcoded paths)

Inspiration

Based on research from HKUDS/ClawWork on economic incentives for AI agents. Adapted for OpenClaw agent systems.

License

MIT

Author

OpenClaw Community

Version

1.0.0