clawcost_skill

Overview

This skill tracks OpenClaw agent spending and provides quick daily, weekly, and all-time cost summaries. It outputs a compact JSON and a tree-style console summary showing balance, today vs budget, weekly totals, model breakdowns, and recent daily costs. Use it to monitor model-level spending and catch budget overruns early.

How this skill works

Run the clawcost script with an optional --budget or --set-balance flag. The script inspects usage logs and aggregates costs into fields: balance (initial, spent, remaining), today (cost, budget, pct), week, total (cost and tokens), models (all-time breakdown), models_today, and daily (last 7 days). It prints JSON for automation and a human-friendly tree summary with alerts when thresholds are exceeded.

When to use it

• Daily check-in to ensure agents stay within a daily budget.
• Weekly reviews to compare model-level spend and spot expensive models.
• After topping up funds — set initial balance with --set-balance.
• When you suspect a runaway agent or sudden cost spike.
• Before deploying new agents to confirm available balance.

Best practices

• Set an initial balance using --set-balance so remaining funds calculate automatically.
• Provide a sensible daily budget with --budget to enable meaningful pct alerts.
• Automate a daily cron job to capture trends and save the JSON output.
• Skip $0 models in summaries to keep output focused on actual spenders.
• When over budget, consider switching traffic to a cheaper model (e.g., Haiku).

Example use cases

• Run python3 scripts/clawcost.py --budget 10 each morning to get today's pct and model hot-spots.
• Top up funds with python3 scripts/clawcost.py --set-balance 50.00 and verify balance remaining.
• Integrate JSON output into a dashboard that triggers alerts when today.pct > 80%.
• Investigate a sudden cost increase by reviewing models_today and the last 7 daily entries.
• Archive historical outputs to compare month-over-month agent spending patterns.