Rome

Rome describes itself as an agentic OS for humans and agents. Instead of scaling only models, it focuses on the environment an agent works in: tools, workflows, memory and interfaces. It is a guardrailed space where people and agents collaborate, agents build their own harnesses and standard procedures under guidance, and capabilities that prove themselves are kept so later work builds on them.

It also positions itself as an open-source alternative to Grok Bot and Meta's Muse, with persistent agents, scheduled tasks and custom apps. You can run it yourself with a single Docker script that pulls the published image and starts a dashboard bound to localhost on port 7663, with state kept in named Docker volumes so upgrades preserve data. First-run onboarding is open to whoever reaches it first, so exposing it beyond the machine requires an explicit bind option. Telemetry export is off unless an OpenTelemetry endpoint is set.

Rome Cloud, currently a preview, provisions a private Rome environment for each user. The code is written in TypeScript and licensed under MIT, and the topics reference Claude Code and Codex alongside self-hosting and workflow automation.