A comprehensive toolkit designed to drastically reduce token usage and API expenditures through intelligent context management and task-based model routing.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install token-optimizer-qsmtco
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-optimizer-qsmtco 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 Token Optimizer?
The Token Optimizer is a specialized suite of utilities for developers looking to scale their AI agent deployments without incurring massive costs. By addressing the primary drivers of token bloat—such as excessive context loading and inefficient model selection—this skill provides the necessary scripts to automate cost-saving measures. It is a critical component for managing professional-grade Openclaw Skills where budget tracking and performance optimization are paramount.
This skill focuses on local file analysis to categorize user intent and system requirements, ensuring that every interaction uses the most cost-effective path. Whether you are running a single personal agent or a fleet of managed agents, the Token Optimizer provides the logic to swap expensive reasoning models for faster, cheaper alternatives when appropriate.
Token Optimizer Use Cases
- Scaling multi-agent deployments where high-frequency heartbeats lead to prohibitive API costs.
- Minimizing context window usage by dynamically loading only relevant files for specific user prompts.
- Enforcing daily budget caps for Openclaw Skills to prevent unexpected billing spikes.
- Transitioning background maintenance tasks and cronjobs to free or low-cost model tiers.
How Token Optimizer Works
- The context optimizer analyzes the complexity of the user's prompt to recommend a minimal set of required files instead of loading the entire workspace.
- A model routing script classifies the task into tiers, forcing simple communication or background tasks to use low-cost Quick models.
- The heartbeat manager tracks the last execution time of periodic checks, enforcing intervals and quiet hours to skip unnecessary API calls.
- Real-time usage tracking compares current consumption against a pricing database to alert users or switch providers when daily limits are reached.
Token Optimizer Setup
Install the toolkit by preparing your workspace and generating optimized configurations:
# Generate an optimized AGENTS.md with lazy loading support
python3 scripts/context_optimizer.py generate-agents
# Deploy the optimized heartbeat template to your workspace directory
cp assets/HEARTBEAT.template.md ~/.openclaw/workspace/HEARTBEAT.md
# Refresh live model pricing from OpenRouter (optional)
export OPENROUTER_API_KEY="your_key_here"
python3 scripts/token_tracker.py refresh-pricing
Token Optimizer Data Schema & Taxonomy
The skill utilizes several local files to manage state and optimization logic for Openclaw Skills:
| File | Purpose |
|---|---|
pricing.json |
A local database of model costs used for budget forecasting and routing. |
HEARTBEAT.md |
An optimized instruction file that governs periodic agent checks. |
AGENTS.md.optimized |
A template for agent definitions that implements conditional context loading. |
scripts/ |
Contains the Python logic for context analysis, routing, and tracking. |
Token Optimizer Advanced Features
- Intelligent task routing that identifies greetings and acknowledgments to prevent the use of high-tier models for casual chat.
- Dynamic multi-provider fallback strategies using a comprehensive guide for OpenRouter, Together.ai, and Google AI Studio.
- Automated cronjob optimization that identifies background tasks suitable for free model tiers.
- Context level recommendations that can reduce initial prompt tokens by up to 80% through selective file loading.
SKILL.md
Loading
Related Openclaw Skills
Session State Tracker
A robust lifecycle management tool that prevents context loss by automatically persisting and restoring AI agent state across compactions and restarts.
Curated Search
A domain-restricted local search engine that provides authoritative technical documentation results without web spam.
ClawShop
An API-driven skill for managing product listings and metadata on Taobao and Goofish marketplaces.
Monkeytype Tracker
A powerful typing performance monitor that integrates Monkeytype statistics and personalized improvement tips into your AI agent workflow.
tokenQrusher
tokenQrusher is a high-performance optimization system that drastically reduces AI agent token usage through intelligent context filtering and heartbeat management.
YaCy Skill
Manage and integrate a local, decentralized YaCy search engine into your AI workflows for private web searches.
Featured*
