An automated model selection engine that routes AI requests based on complexity analysis and budget constraints to maximize cost efficiency.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install openclaw-smart-router
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 openclaw-smart-router 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 Smart Router?
The Smart Router is a powerful addition to the ecosystem of Openclaw Skills, designed to eliminate wasteful spending on high-end LLMs for trivial tasks. By acting as an intelligent middleware, it intercepts every request and performs a real-time complexity analysis, scoring tasks from 0.0 to 1.0. This allows the system to intelligently route simple requests to lightweight models like Haiku or GPT-3.5, while reserving premium models like Opus or GPT-4 for expert-level reasoning and complex debugging.
Beyond simple routing, this skill incorporates a sophisticated pattern learning engine that adapts to your specific development style. It tracks which models successfully resolve specific types of tasks—such as refactoring or unit testing—and refines its selection logic over time. For developers looking to scale their AI-driven workflows, the Smart Router provides a transparent way to manage budgets, track ROI, and ensure peak performance without manual intervention.
Smart Router Use Cases
- Automatically reduce LLM API expenditures by 30-50% without sacrificing quality.
- Implement autonomous budget-aware routing to prevent unexpected overages during heavy development.
- Transition from manual model selection to an automated, data-driven workflow.
- Optimize agent performance by leveraging pattern learning to match tasks with the most successful models.
- Monitor real-time cost savings and model distribution via a local web dashboard.
How Smart Router Works
- Intercepts the request by hooking into the OpenClaw API lifecycle before the call is sent to a provider.
- Performs a complexity analysis by evaluating context length, code presence, and task specificity.
- Checks the user-defined budget and daily spending limits to ensure the request is within financial bounds.
- Selects the optimal model based on the complexity score (e.g., Simple tasks go to Haiku, Expert tasks go to Opus).
- Routes the request to the selected provider (Anthropic, OpenAI, Google, etc.) and executes the task.
- Learns from the output by tracking success rates and updating internal routing patterns for future requests.
Smart Router Setup
To install the Smart Router and start optimizing your Openclaw Skills workflow, use the following command:
claw skill install openclaw-smart-router
You can then view your routing statistics or open the local management dashboard:
# View current routing and savings stats
claw router stats
# Open the analytics dashboard
claw router dashboard
Smart Router Data Schema & Taxonomy
The skill maintains a local-first data architecture to protect privacy and minimize latency. All data is stored in the ~/.openclaw/openclaw-smart-router/ directory.
| Component | Data Type | Purpose |
|---|---|---|
history.json |
Log | Stores every routing decision, complexity score, and associated cost. |
patterns.db |
Database | Contains learned associations between task types and model success rates. |
config.json |
Settings | Defines budget limits, preferred providers, and scoring thresholds. |
stats/ |
Analytics | Pre-computed ROI and model distribution metrics for the dashboard. |
Smart Router Advanced Features
- Autonomous x402 Payments: Enables agents to self-fund a Pro subscription if the projected savings exceed the 0.5 USDT/month cost.
- Multi-Provider Load Balancing: Seamlessly switches between Anthropic, OpenAI, and Google Gemini based on real-time availability and cost.
- Deep Pattern Learning: Identifies specific task signatures, such as 'Python debugging', to prioritize models that have historically performed best.
- Efficiency Suite Integration: Combines with the Context Optimizer and Memory System to achieve a total cost reduction of up to 80%.
- Local Web Dashboard: Provides a comprehensive UI at
http://localhost:9093for tracking complexity distribution and model performance.
SKILL.md
Loading
Related Openclaw Skills
OpenClaw Memory
A persistent memory layer that enables AI agents to remember facts, preferences, and context across sessions using semantic search and autonomous payments.
OpenClaw Ledger
A tamper-evident audit trail that records and verifies every workspace change in a hash-chained log for AI agents.
Context Optimizer
An intelligent context compression engine that reduces token costs by up to 60% through pruning, deduplication, and summarization.
OpenClaw Security Action
A specialized GitHub Action that automatically scans agent workspaces and Openclaw Skills for security vulnerabilities like exposed secrets and prompt injection.
Craft.do Integration Skill
A comprehensive REST API integration for Craft.do that enables programmatic document management, task automation, and folder organization.
Decision Memo Writer
Turn information overload from long reports and complex threads into concise, decision-ready memos with clearly defined risks and next steps.
Featured*
