claude-usage

What This Skill Does

The claude-usage skill is a specialized diagnostic tool designed for OpenClaw users to gain granular visibility into their Claude API consumption. Because Claude subscription limits are notoriously opaque, this skill reverses the credit-based consumption model to provide a clear, real-time dashboard of your usage. It tracks input, output, and cache-write tokens against your specific subscription tier (Pro, 5x, or 20x), translating technical token metrics into actionable financial and capacity insights. Whether you are worried about hitting your 5-hour rate limit or curious about which specific agent session is burning through your budget, this utility provides the transparency needed to manage your AI workspace effectively.

Installation

To install this skill, use the clawhub CLI within your OpenClaw environment:

clawhub install openclaw/skills/skills/lemodigital/claude-usage

After installation, you must initialize your configuration to match your account settings. Navigate to your Claude Settings page, identify your weekly reset time, and run:

python3 {SKILL_DIR}/scripts/claude-usage.py "2026-02-09 14:00" --plan 5x --save

This creates a persistent configuration file, allowing you to run quick queries thereafter without re-specifying your plan or timezone.

Use Cases

This skill is indispensable for power users who leverage OpenClaw to manage high-volume AI tasks. Use it to:

• Monitor usage trends: See if your average consumption is trending toward a plan upgrade.
• Identify resource-heavy tasks: Pinpoint which sessions or automated workflows are the primary contributors to your credit depletion.
• Rate-limit management: Receive immediate warnings if you are approaching the sliding 5-hour ceiling, allowing you to pause non-essential background tasks before being throttled.
• Audit performance: Analyze the token-efficiency of different models (Opus vs. Sonnet) for your specific project needs.

Example Prompts

1. "OpenClaw, check my current Claude budget. How close am I to my 5-hour limit?"
2. "Show me the last 5 sessions, ranked by the most credits consumed."
3. "Summarize my usage for the 'research-main' session and break down the token costs."

Tips & Limitations

• Cache Efficiency: Remember that cache reads are free; focus your prompt engineering on maximizing cache usage to lower costs.
• Scope: This tool only analyzes OpenClaw session files stored in ~/.openclaw/agents/main/sessions/. It cannot track usage from the Claude web interface directly.
• Accuracy: Since this relies on reverse-engineered credit formulas, minor discrepancies compared to official billing may occur due to rounding differences in provider APIs.