rome-os/rome

Rome is pitching an open-source, self-hostable agent environment where repeated work becomes reusable apps, actions, skills, and persistent data.

The repo describes Rome as a guardrailed workspace for humans and agents, with Docker quickstart and a preview Rome Cloud option. Its core claim is that agents should compound through the environment around them: git-tracked capabilities, app-private databases, workflows, memory, and purpose-built interfaces. Rome Apps are presented as persistent products rather than prompt wrappers, meant for work like code review loops, inbox triage, price tracking, and scheduled briefs. GitHub · LLM repos' note