Clawd Throttle

Description

Automatically routes LLM requests to the cheapest capable model based on prompt complexity. Classifies prompts on 8 dimensions in under 1ms, supports 8 providers and 27+ models across three routing modes (eco, standard, performance). Works as both an MCP server and an HTTP reverse proxy compatible with any OpenAI/Anthropic client.

Install

npx clawhub@latest install clawd-throttle

Click to copy the install command