justlucyhq/lucy

Lucy is a self hosted AI chat platform built for teams who want to use AI tools without sending their company data to someone else's cloud. It connects to OpenAI, Anthropic Claude, Google Gemini, and several other providers, as well as local models run through Ollama or LM Studio, all through one interface where you can switch models in the middle of a conversation and watch responses stream in as they are generated. Chat history and settings can be kept entirely in the browser for a zero infrastructure standalone mode, or saved to a Supabase PostgreSQL database when a team needs multiple users and access across devices. In connected mode, Lucy also builds a memory of facts, preferences, and past conversations that carries across sessions, using a mix of search techniques to pull back only what is relevant to the current message, along with a privacy guard meant to avoid storing secrets. A lexical fallback version of memory works even in standalone mode without a database. Beyond chat, Lucy includes a visual workflow builder, a drag and drop canvas where you can wire together multi step AI pipelines from building blocks like an AI agent step, a knowledge base lookup, conditions, filters, and HTTP calls. In connected mode these workflows run on the server and keep going even if you close the browser tab, can be triggered by a schedule, a webhook, or a data change, and will retry automatically if a step fails. There is also a terminal command line client for chatting from any shell, and an integration layer that lets other applications register with Lucy so it can read their data as context and trigger actions on their behalf, plus an embeddable chat widget that can be added to any web page with one script tag. The project is written in TypeScript with Next.js, React, and Tailwind CSS on the frontend, with Supabase as the optional backing database. It is released under the AGPL-3.0 license. The full README is longer than what was shown.