Alice is an advanced UI for interacting with AI in a personalized way. It lets you customize everything you need, including connections to popular API providers and even open-source models that work locally on your machine. By default, all settings and conversation history remain on your device.
The app focuses on delivering a high-quality interface and cutting-edge multi-agent logic. It includes integrations with external tools through Model Context Protocol (MCP), which works in two modes: native mode and code mode, which runs on an embedded Node.js runtime.
Alice UI is designed to be minimalistic, clean, and customizable. At first glance, it is a simple chat interface. In practice, it covers image generation and editing, artifacts, code interpreter, multi-agent workflows, agentic memory, and working with text files.
Story
Alice is my side project. I’ve been working on it since December 2022, right after the “ChatGPT moment”, and its main goal is to serve as a sandbox in which I explore what generative AI can and cannot do.
Developing this app taught me the core principles of interacting with LLM APIs, managing context, and building architecture for a multi-agent system flexible enough for users to set up their own system prompts and connect their own tools.
There is also one additional challenge related to building Alice: its core logic is written in Rust, which at the time was an entirely new language for me.