The Problem Every AI coding tool loads 15K-28K tokens of system prompts before you can even ask a question. The AI spends most of its attention reading the manual, not solving your code. The Solution Huiyu Pi strips everything non-essential: - ~80 tokens system prompt (not 20K) - ~0.3s first token response - 90%+ cheaper Token - 4 basic tools only Quick Start npx huiyu-pi Built on Huiyu Pi is built on pi (by earendil-works) and pi-forge www.huiyu.ai, https://github.com/huiyu9144/Huiyu-Pi

Hey Product Hunt! 👋 I built Huiyu Pi because I was frustrated with the overhead in existing AI coding tools. Every request carried 20K+ tokens of context that the AI had to parse before doing any actual work. That's like paying a consultant to read the entire company handbook before answering "what's 2+2?" So I stripped everything non-essential: - ~80 tokens system prompt (not 20K) - ~0.3s first token response - 90%+ cost reduction The core idea is "less context = better AI focus." It's fully self-hosted — your code, API keys, and conversations never leave your machine. MIT licensed, open source. Try it: npx huiyu-pi I'd love to hear your feedback! What do you think of the minimal context approach?

Huiyu Pi

Self-hosted AI coding agent with ~80 tokens system prompt

Self-hosted AI coding agent with ~80 tokens system prompt