Tell us what needs to run
Share the runtime, expected usage, current hosting pain and budget. Don't send credentials, proprietary datasets or your customers' personal information.
Managed compute / Pilot inquiries open
Your workload.
A managed place
to run it.
For agencies and developers with a working application, automation worker or build pipeline—and no appetite to manage another server.
Qualification first. No payment is collected. Availability is confirmed individually.
The right work
Keep the app.
Hand off the host.
You bring a working workload. We agree on the environment and operating responsibilities before anything is deployed.
- 01
Background automation
Scheduled catalog updates, report generation and internal integration workers. Bring your own application and authorized API accounts.
- 02
Staging & internal applications
A place for an agency's staging app or an internal tool where single-host downtime is acceptable. Not a high-availability production platform.
- 03
Private CPU workers
Build jobs, test suites and restartable batch processing. Workloads are reviewed for CPU, memory, disk and traffic requirements.
Scope before checkout
Proposed pilot scope
These are discussion starting points, not currently provisionable products. Final resources, onboarding, traffic limits and operating terms require a written quote.
| Environment | Managed Worker | Managed Team |
|---|---|---|
| Proposed price | $299 / month | $599 / month |
| Linux virtual machine | 4 vCPU / 8 GiB RAM | 8 vCPU / 16 GiB RAM |
| Storage allocation | 80 GiB | 160 GiB |
| Agreed applications | 1 application or worker | Up to 2 related workloads |
| Operating scope | Setup of an agreed deployment, scheduled OS maintenance and service checks | |
| Included operator time | 1 hour / month | 2 hours / month |
| Backup destination | Customer-provided off-host destination; restore test required before activation | |
vCPU means virtual CPU allocation, not dedicated physical cores. The server and network are shared. CPU, storage and traffic policies will be specified before an agreement; no unlimited resources are offered.
Application development, debugging your code, AI model/API usage, paid software, third-party storage charges and work beyond the included operator time are not included. Extra work requires a separate quote and approval.
The pilot uses a single physical host. No GPU, high-availability failover, uptime SLA or 24/7 response commitment is included.
Before a paid pilot
Fit first.
Then a verified setup.
Agree on scope and limits
We check suitability and availability, then put the deployment, support window, traffic allowance, backup responsibilities, price and cancellation process in writing.
Verify before activation
Commercial pilots only begin after the required licensing and provider checks, tenant-isolation tests, resource controls and an off-host restore test pass. An inquiry does not reserve capacity or trigger a charge.
Plain answers
Know the limits.
Is this a VPS subscription I can buy today?
No. This page collects interest in a managed pilot. It is separate from the existing Compute Engines VPS catalog, and no new subscription is created here. Availability and final scope must be confirmed.
Can I run an AI agent here?
We can assess a customer-owned agent runtime as a CPU application. Model access, token charges and authorized third-party accounts remain yours. We are not selling GPU inference, an AI subscription, pooled accounts or guaranteed agent results.
Which workloads are not a fit?
Always-on critical transaction systems, workloads requiring regulated-data controls, latency-sensitive live trade execution, bulk email, proxy resale, mining and abuse-oriented automation are outside this initial pilot. Restartable internal work is the starting point.
What hardware is behind it?
The audited host is a Dell PowerEdge R7625 with two AMD EPYC 9654 processors, 192 physical cores in total and 640 GiB of installed memory. Storage is SATA SSD in RAID 1 mirrors—not NVMe. These are host specifications, not resources dedicated to each customer. The active physical uplink is 1 Gbps shared across the server; it is not a per-customer bandwidth guarantee.
How is my inquiry handled?
The inquiry form stores the details you choose to provide in our existing access-controlled site system so we can evaluate and reply to your request. It does not enroll you in marketing. The form includes the inquiry privacy notice. You can also email Compute Engines support.
What would you stop babysitting?
Tell us about one workload. We'll start with its requirements—not a bigger hosting plan.