clawculator

What This Skill Does

Clawculator is a specialized diagnostic utility designed for OpenClaw users to perform cost forensics and billing anomaly detection. As an agent skill, it provides a deterministic analysis of your workspace, sessions, and configuration files to identify hidden expenses. Unlike many other tools, Clawculator operates entirely offline with zero network connectivity. It requires only the local node binary to execute, ensuring that your sensitive configuration files and session data remain strictly within your local environment. The tool reads your local OpenClaw configuration and session usage metrics to flag potential cost traps, such as running heartbeat services on expensive paid models, inadvertent skill polling loops, or bloated workspace contexts that inflate your token usage.

Installation

To add this skill to your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/echoudhry/clawculator

Once installed, the tool is immediately ready for use without any external configuration or API keys, as it processes data locally using the bundled source code.

Use Cases

Clawculator is best used when you want to audit your monthly operational costs or investigate unexpected spikes in token usage. It is highly effective for developers who want to maintain an efficient, cost-optimized OpenClaw setup. It helps you catch "leaky" configurations, such as high-cost models being used for low-priority tasks like background heartbeats or automated logging hooks. It also allows you to generate shareable snapshots of your environment's health, which can be useful for debugging setup complexity with community members or team leads.

Example Prompts

1. "check my costs"
2. "clawculator snapshot"
3. "analyze spend"

Tips & Limitations

Clawculator is a read-only analyzer by default. If you choose to use the --md flag, it will create a clawculator-report.md file in your directory to give you a permanent record of your findings. Note that while it reads your workspace file counts, it does not read the actual contents of your files, prioritizing privacy and performance. Because it relies on static analysis of your configuration and log files, it cannot predict future usage patterns, but it is excellent at identifying current architectural inefficiencies. Always run the --help command if you need a refresher on custom output paths.