model-usage-linux

What This Skill Does

The model-usage-linux skill serves as a diagnostic and reporting tool designed specifically for Linux users of the OpenClaw AI platform. It provides a robust, command-line interface to analyze local session logs stored in JSONL format. By parsing these session files, the skill calculates detailed metrics, including token consumption (input, output, and cached tokens), turn counts, and estimated financial costs associated with specific LLM models. It is the direct functional equivalent to the legacy macOS-only model-usage or CodexBar tools, ensuring Linux developers maintain parity in tracking their AI infrastructure consumption.

Installation

To integrate this utility into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/hablabechir/model-usage-linux

Once installed, ensure your environment has Python 3 configured. The tool relies on the standard Python interpreter to parse session logs. No additional heavy dependencies are required, making it lightweight for server environments or local development machines. The tool defaults to looking for sessions in your local ~/.openclaw/agents/main/sessions directory, though this path can be customized via the --sessions-dir flag.

Use Cases

This skill is indispensable for power users, developers, and administrators who need to maintain tight control over their AI budgets. Use it to:

• Monitor monthly token expenditure across different project sessions.
• Analyze which models are the most cost-effective for specific tasks (e.g., comparing gpt-4o vs claude-3-5-sonnet).
• Audit usage to detect anomalous high-token-count sessions that might indicate inefficient prompting.
• Export consumption data for automated reporting via the --format json flag.

Example Prompts

1. "How much have I spent on token usage across all my sessions this month?"
2. "Show me a breakdown of which models I've used the most in my development sessions."
3. "Can you provide a summary of my total token consumption and API costs for the last few days?"

Tips & Limitations

• Tip: If you frequently analyze logs, consider aliasing the command in your .bashrc or .zshrc file for faster access.
• Tip: Use the JSON output format to pipe results into visualization tools or custom dashboarding scripts for better data presentation.
• Limitation: This skill is strictly read-only; it does not modify your session files, ensuring data integrity.
• Limitation: It requires access to your local filesystem. Ensure your user permissions allow reading the directory where OpenClaw stores its session history.