Advanced options are for users who want 24/7 availability, stronger isolation, or full infrastructure control. These options run your assistant on a remote machine, either managed by Vellum or managed by you.
Status: These options are not generally available yet. They're on the roadmap and actively in development.
Your assistant runs in Vellum's cloud infrastructure. We handle compute, networking, storage, and uptime. This is the easiest option and the one we recommend for most users.
The tradeoff is privacy: your data lives on and passes through Vellum's servers. If keeping your data off third-party infrastructure is a hard requirement, consider local hosting or User-Hosted Remote instead.
A common question: if my assistant is running in the cloud, can it still interact with my Mac? Yes.
The distinction is between tools that run on the assistant's computer vs. tools that run on yours:
The same pattern will apply to browser use, computer use, and file access. The assistant can go wild on its own machine in the cloud, and it can also reach back to your computer, but only when you're on your computer and explicitly grant permission.
If your laptop is closed, the assistant can still do its own work (run schedules, process emails, research things) but it won't be able to open files on your Mac or use your local tools until you're back.
You provide the infrastructure, Vellum provides the software. Your assistant runs on machines you own and control: a cloud VM, a dedicated server, or even a Mac Mini at home.
Deploy your assistant to your own Google Cloud project. You manage IAM, networking, and compute. Your data stays in your GCP account.
vellum hatch --remote gcpBest for users or teams already on GCP who want to keep everything within their existing cloud setup.
Deploy to your own AWS account. Same idea as GCP: you own the infrastructure, credentials, and data.
Best for users or organizations standardized on AWS identity, networking, and observability.
A dedicated Mac Mini running at home or in an office, with the assistant running locally on it (native, Docker, or Apple Container). You connect to it from your primary computer through the Vellum desktop app.
This gives you 24/7 availability, full data ownership on hardware you physically possess, and Apple Container isolation. The tradeoff: your files and data live on the Mac Mini, not your daily laptop. Accessing your laptop's files requires the same host_bash tunneling as any remote option.
All remote options (Vellum Managed and User-Hosted) share the same fundamental tradeoff compared to local hosting:
We're building features like host_bash, host browser use, and host computer use to bridge this gap. The goal: the assistant can still interact with your personal machine when you're available, and do everything else autonomously on its own machine when you're not.
| Option | Managed by | Data location | Setup effort | 24/7 available |
|---|---|---|---|---|
| Vellum Managed | Vellum | Vellum's cloud | Minimal | Yes |
| GCP | You | Your GCP project | Significant | Yes |
| AWS | You | Your AWS account | Significant | Yes |
| Mac Mini | You | Your hardware | Moderate | Yes |