An OS where AI agents leave behind real code
Rome is a self-hostable TypeScript environment that turns recurring agent tasks into persistent, git-tracked apps with their own interfaces and databases.

What it does
Rome is a TypeScript-based “agentic OS” that runs persistent AI agents alongside humans in a guardrailed environment. Instead of treating every request as a fresh chat, it lets agents build reusable workflows, design their own SOPs, and orchestrate scheduled tasks. Proven capabilities accumulate as executable code, memory, and purpose-built interfaces that are meant to raise the ceiling for future work.
The interesting bit
The project treats apps as first-class capabilities—not thin prompt wrappers, but installable products with their own databases, web UIs, and agent-owned actions that other agents can later discover and reuse. Rome can scaffold these apps from a plain-language description, producing ordinary git-tracked source code rather than hidden model state. It is explicitly positioned as an open-source, self-hosted alternative to Grok Bot and Meta’s Muse.
Key highlights
- Self-hosted or cloud-preview runtime for persistent agents, with state surviving in Docker volumes
- Apps ship via an
app.yamlmanifest and can bundle actions, agents, skills, hooks, web UIs, and private databases - Actions are typed, reusable building blocks that agents, routines, and interfaces call—able to pause for human approval
- Built-in comparison to Grok Bot, Muse, Hermes, and Manus emphasizes executable software persistence over mere text memory
- Scaffolds new apps from natural language descriptions, iterating inside the same conversation
Caveats
- Rome Cloud is still a preview, and the repository’s development stack is currently framed as the contributor path, not a final production self-hosting distribution
- At 502 stars, the ecosystem is early; the App Store and broader community reuse are still growing
- The README truncates mid-comparison with Wabi, suggesting some documentation is unfinished
Verdict
Worth a look if you want agents that leave behind real software—git-tracked, self-hosted, and composable—rather than vanishing into chat history. Skip it if you need a polished, turnkey consumer product today; Rome is still very much a build-your-own-environment affair.
Frequently asked
- What is rome-os/rome?
- Rome is a self-hostable TypeScript environment that turns recurring agent tasks into persistent, git-tracked apps with their own interfaces and databases.
- Is rome open source?
- Yes — rome-os/rome is open source, released under the MIT license.
- What language is rome written in?
- rome-os/rome is primarily written in TypeScript.
- How popular is rome?
- rome-os/rome has 502 stars on GitHub.
- Where can I find rome?
- rome-os/rome is on GitHub at https://github.com/rome-os/rome.