Aivory
A self-hosted AI workspace: multi-model chat, multi-step tool calls, a persistent Python sandbox, retrieval over your own documents and Deep Research, for individuals or teams.
AWhat it is
Aivory is an AI workspace you host yourself. Conversations, research, creation and team knowledge live in one place, and both the product and the data stay in your hands. It works with Claude, GPT, Gemini and any OpenAI-compatible provider.
BWhat it does
Switch models by task and keep conversations organised in projects; search, read web pages and documents, then turn the results into code, images, reports, spreadsheets or slides; build libraries from your team’s documents and cite them in conversation. Personal and team workspaces stay separate, and admins manage model channels, members, usage and backups in one place.
CHow it’s built
A Go backend with a React and TypeScript frontend, deployed with Docker Compose. Tool calls can chain over several steps, and code runs in a persistent Python sandbox. The full profile is for teams; the personal profile runs on a single SQLite database on one server.
DWhere it is
Open source under Apache 2.0, with its own documentation site and a live demo to try before you deploy it.
EThe lines that matter
git clone https://github.com/hjxwz123/Aivory.gitcd Aivory/deploycp .env.personal.example .env.personal # then set JWT_SECRETdocker compose --env-file .env.personal -f docker-compose.personal.yml up -d