Switchboard optimizes AI agent operational costs by intelligently routing tasks to different model tiers based on their complexity and requirements.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install 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 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 Switchboard?
Switchboard is a sophisticated routing layer designed to maximize efficiency for AI agents. By recognizing that approximately 80% of agentic tasks are routine operations like file reading or formatting, this skill prevents overspending on premium models for simple chores. It leverages OpenRouter to provide access to multiple model tiers, from zero-cost options to high-reasoning frontier models.
Integrating this logic into Openclaw Skills allows developers to maintain high performance while reducing monthly API expenditures by up to 10x through a strategic model hierarchy. It ensures that expensive reasoning power is reserved for complex debugging and architectural decisions, while janitorial tasks are handled by faster, cheaper alternatives.
Switchboard Use Cases
- Deciding which model to use for a specific sub-task based on cost and reasoning needs.
- Spawning sub-agents for high-volume background tasks like URL fetching or status checks.
- Automatically escalating to a premium reasoning model when a standard model fails a task.
- Scaling agent operations without exponential increases in API costs.
How Switchboard Works
- The skill analyzes the task complexity using classification signals like debug, summarize, or read.
- It applies hard constraints such as tool support requirements and estimated token length to filter available models.
- A decision algorithm selects the cheapest model within the appropriate tier (Free, Cheap, Mid, or Premium) that meets the task criteria.
- The task is executed via the main session or a spawned sub-agent using the selected model via OpenRouter integration within Openclaw Skills.
Switchboard Setup
To use Switchboard, you must configure your OpenRouter API key. Add it to your configuration file:
# Edit ~/.openclaw/openclaw.json
{
"openrouter_api_key": "sk-or-v1-..."
}
You can also set a default session model in your config.yml file:
# config.yml
model: anthropic/claude-sonnet-4
Switchboard Data Schema & Taxonomy
Switchboard organizes models into a four-tier hierarchy based on pricing and capability for use with Openclaw Skills:
| Tier | Category | Pricing Range | Primary Use Case |
|---|---|---|---|
| Tier 0 | Free | $0.00 | Non-critical background tasks |
| Tier 1 | Cheap | $0.02 - $0.50/M | Routine agent loops and file operations |
| Tier 2 | Mid | $1.00 - $5.00/M | General interactive work and code generation |
| Tier 3 | Premium | $5.00+/M | Complex reasoning and deep debugging |
Switchboard Advanced Features
- Dynamic mid-session model switching using the /model command to adapt to changing task difficulty.
- Model-specific sub-agent spawning for granular task management and cost control.
- Latency-aware routing for real-time agent loops where response speed is critical.
- Automatic escalation triggers that switch to Tier 3 models when reasoning attempts on lower tiers fail.
SKILL.md
Loading
Related Openclaw Skills
Tick Multi-Agent Coordination
A Git-backed Markdown protocol that provides a transparent, machine-parseable framework for coordinating complex tasks between humans and AI agents.
memory-tools
An agent-controlled memory plugin that provides persistent, file-based knowledge storage with semantic search for Openclaw Skills.
Kalshi
The Kalshi skill for Openclaw Skills enables automated trading and market analysis on the first CFTC-regulated prediction market exchange.
Xobni.ai Email Skill
Provide AI agents with real-world email addresses, semantic search capabilities, and MCP server integration for seamless communication.
flomo-add
A streamlined utility to instantly capture and send notes to your flomo account using webhook automation.
flomo-sync
A Python-based utility designed to incrementally synchronize and backup flomo memos into locally stored, formatted Markdown files.
Featured*
