clawbicle
Smart openclaw agent router that automatically routes the tasks to agents based on tasks.
Core trick
Uses MAP-Elites search over cost, quality and latency to pick an LLM for each prompt.
Tags
Description
Smart openclaw agent router that automatically routes the tasks to agents based on tasks. (didnt have enough time for demo :( so heres a fun link instead!
README excerpt
<p align="center"> <img src="assets/openclaw-tui.png" alt="OpenClaw Router TUI" width="100%"> </p> <h1 align="center">OpenClaw Router</h1> <p align="center">Intelligent LLM routing with MAP-Elites optimization. Routes prompts to the optimal model based on cost, quality, and latency tradeoffs.</p> --- ## Cost Reduction Results Simulated workload: 500 calls/day × 30 days = 15,000 calls/month Avg input tokens per call: 800 ### Monthly Cost Comparison | Strategy | Monthly Cost | Avg Quality | Avg Latency | |----------|-------------|-------------|-------------| | **Router [min-cost]** | **$89.04** | **84.5%** | **260ms** | | **Router [balanced]** | $147.82 | 87.2% | 385ms | | **Router [low-latency]** | $198.34 | 89.1% | 180ms | | Gemini 2.0 Flash Lite only | $135.00 | 65.0% | 100ms | | GPT-4.1 Nano only | $180.00 | 72.0% | 100ms | | Gemini 2.0 Flash only | $180.00 | 78.0% | 150ms | | Gemini 2.5 Flash only | $270.00 | 88.0% | 200ms | | GPT-4.1 Mini only | $360.00 | 85.0% | 250ms | | Llama 4 Scout only | $405.00 | 80.0% | 200ms | | Mistral Small only | $450.00 | 78.0% | 120ms | | DeepSeek V3 only | $495.00 | 87.0% | 400ms | | Llama 4 Maverick only | $720.00 | 88.0% | 350ms | | GPT-4.1 only | $720.00 | 93.0% | 600ms | | Claude Haiku 4 only | $1,800.00 | 86.0% | 180ms | | Mistral Large only | $2,160.00 | 89.0% | 500ms | | o3 (Reasoning) only | $2,700.00 | 95.0% | 3000ms | | Gemini 2.5 Pro only | $5,400.00 | 94.0% | 800ms | | Claude Sonnet 4 only | $5,400.00 | 93.0% | 500ms