Show HN: Potluck – Run local AI across the computers you own

Category: ai-ml

Tags: local-inference, mesh-networking, llama-cpp, wireguard, mcp, openai-api-compatible

Score: 8.0/10 (Innovation: 8, Technical: 9, Documentation: 8, Utility: 7)

Potluck is a local-first AI runtime that combines llama.cpp inference, WireGuard mesh networking, and per-request routing scopes to run models across the computers you own. It offers desktop and CLI interfaces, an OpenAI/Anthropic-compatible API, MCP memory server, and even split-model execution across machines, making distributed local inference accessible.

Target audience: AI engineers, backend developers, devops, privacy-conscious users, homelab enthusiasts

Repository: https://trypotluck.ai/ · Python · MIT · 2 stars

View on Hacker News