A Linux-native tool to analyze AI model token consumption and financial costs from session logs.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install model-usage-linux
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install model-usage-linux using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Model Usage (Linux)?
Model Usage (Linux) is a specialized utility designed for developers and power users working with the Openclaw Skills ecosystem. It serves as a robust Linux alternative to macOS-specific tracking tools, allowing users to deep-dive into their AI agent sessions. By parsing local JSONL session logs, it provides clear visibility into model performance and resource consumption, ensuring you stay within budget while optimizing your workflows.
Model Usage (Linux) Use Cases
- Monitoring total API expenditure across multiple LLM providers.
- Identifying which AI models are consuming the most tokens in your Openclaw Skills setup.
- Generating usage reports for billing or project resource allocation.
- Debugging session costs by analyzing input, output, and cache token metrics.
How Model Usage (Linux) Works
- The script accesses the local session directory where Openclaw Skills store interaction history in JSONL format.
- It parses each session file to extract metadata, model identifiers, and token counts.
- The tool calculates costs based on predefined model pricing for input, output, and caching.
- It aggregates the data into a summarized view, categorized by model type and interaction turns.
- Results are displayed in the terminal or exported as JSON for further integration.
Model Usage (Linux) Setup
Navigate to your skill directory and run the following command to see a usage summary:
python3 {baseDir}/scripts/usage.py
To output the results in JSON format for automated reporting or integration with other Openclaw Skills:
python3 {baseDir}/scripts/usage.py --format json
If your session files are stored in a non-standard location, use the directory flag:
python3 {baseDir}/scripts/usage.py --sessions-dir ~/.openclaw/agents/main/sessions
Model Usage (Linux) Data Schema & Taxonomy
| Metric | Description |
|---|---|
| Turns | Number of assistant responses per session |
| Input Tokens | Tokens sent to the model in requests |
| Output Tokens | Tokens generated by the model in replies |
| Cache Read/Write | Metrics for optimized token reuse and caching |
| Cost (USD) | Estimated financial cost based on current model rates |
Model Usage (Linux) Advanced Features
- Support for custom session directory paths to manage multiple agent profiles.
- JSON output formatting for seamless integration with external dashboard tools or scripts.
- Granular breakdown of cache-related token savings to evaluate model efficiency.
- Linux-optimized performance for handling large volumes of Openclaw Skills session data.
SKILL.md
Loading
Related Openclaw Skills
Habit AI Health & Nutrition Tracker
A comprehensive health and nutrition tracking skill that leverages the Habit AI API to log meals, water, steps, and mental wellness.
Solana Development Skill
An opinionated end-to-end playbook for building, testing, and deploying high-performance Solana programs and React-based decentralized applications.
Hosting Meta-Skill
A zero-friction deployment orchestrator that syncs local codebases to production URLs via GitHub, Vercel, and Netlify.
VPN & Secure Tunneling Meta-Skill
An orchestration meta-skill for establishing secure, verified network tunnels and resuming blocked workflows via automated VPN management.
Bailian TTS
A high-performance text-to-speech skill leveraging Aliyun Bailian for natural AI voice generation via command-line tools.
Subagent Orchestrator
A sophisticated workflow controller for managing continuous, multi-step AI subagent task lines without manual user re-prompting.
Featured*
