TokenWatch is a comprehensive spending monitor that helps developers track and optimize AI token costs across multiple providers in real-time.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install token-watch
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 token-watch 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 TokenWatch?
TokenWatch is a specialized utility designed for developers who need to maintain strict control over their generative AI expenditures. As one of the most practical Openclaw Skills, it provides a local-first, zero-dependency solution for recording API usage from providers like Anthropic, OpenAI, Google, and Mistral. By automating the calculation of costs based on the latest February 2026 pricing models, it ensures that users are never surprised by their monthly bills.
This skill functions entirely on your local machine, requiring no external API keys or telemetry. It allows users to gain full visibility into where their budget is going and provides actionable suggestions for reducing spend by comparing 41 different models across 10 major providers. It is an essential addition to any privacy-conscious development workflow seeking to maximize efficiency while minimizing overhead.
TokenWatch Use Cases
- Monitoring real-time AI spending across multiple providers to avoid unexpected billing.
- Comparing model costs before execution to select the most budget-friendly LLM for a specific task.
- Setting granular daily or monthly budgets with automated alerts to prevent cost overruns.
- Identifying optimization opportunities where expensive models can be replaced with more efficient alternatives.
- Generating detailed financial reports for AI-integrated projects and team audits.
How TokenWatch Works
- The user initializes the monitor which creates a local storage directory for tracking usage data.
- API usage is recorded either manually or automatically through provided helper functions for popular SDKs like OpenAI and Anthropic.
- The system calculates costs in real-time using an internal database of 41 different model pricing structures updated for 2026.
- Budget thresholds are monitored continuously, triggering automated alerts when spending reaches user-defined percentages.
- Detailed dashboards and optimization reports are generated to provide actionable insights into spending patterns and potential savings.
TokenWatch Setup
Install the skill into your local environment and initialize it within your Python project:
from tokenwatch import TokenWatch
monitor = TokenWatch()
To set up budgets and alerts:
monitor.set_budget(
daily_usd=1.00,
monthly_usd=15.00,
alert_at_percent=80.0
)
TokenWatch Data Schema & Taxonomy
The skill organizes its data locally to ensure privacy and fast access. The following data structures are used:
| Component | Description | Format |
|---|---|---|
| Usage Logs | Stores model name, token counts, task labels, and calculated USD cost. | JSON |
| Budget Config | Stores daily, weekly, and monthly limits alongside alert thresholds. | JSON |
| Pricing Table | A dictionary of 41 models across 10 providers with input/output rates per 1M tokens. | Internal Object |
| Export Reports | Aggregated spending data for specific time periods (today, week, month). | JSON |
TokenWatch Advanced Features
- Automatic recording hooks for Anthropic and OpenAI SDK response objects for seamless integration.
- Intelligent optimization engine that suggests model swaps (e.g., swapping Claude Opus for Sonnet) to save money.
- High-resolution spending dashboard with visual progress bars for budget tracking.
- Support for custom model pricing via simple updates to the internal provider dictionary.
- Multi-period aggregation that allows for deep dives into spending by provider, model, or specific task label.
SKILL.md
Loading
Related Openclaw Skills
Prompt Performance Tester
A model-agnostic benchmarking tool that compares performance, cost, and quality across nine major AI providers simultaneously.
PersistentMind
A persistent, searchable memory layer that prevents AI agents from forgetting user preferences and project context between sessions.
Adaptive Learning Agent
A zero-dependency Python tool for AI agents to capture, store, and retrieve real-time learnings and error resolutions.
Multilingual Greetings
A specialized skill for AI agents to deliver culturally appropriate, multilingual greetings across formal and informal contexts.
Job Auto-Apply Skill
Automate job searches and application submissions across multiple platforms like LinkedIn and Indeed using Clawdbot.
Web Accessibility (WCAG 2.1 AA)
A comprehensive developer framework for building WCAG 2.1 AA compliant websites using semantic HTML, ARIA, and keyboard-first navigation.
Featured*
