SSH
Best for command lines, scripts, runners, builds, log viewing, file synchronization, and process management. Use keys for automation and limit key distribution.
Configuration, Billing & Connectivity
Get the key Cloud Mac details before you buy
VMKeep provides dedicated Apple Silicon physical machines for continuous builds, automation, remote graphics work, and MLX inference. Compare three configurations, five locations, full pricing, and usage limits before ordering.
- 3 tiers
- Fixed configurations
- 5
- Available locations or facilities
- 4
- Daily, weekly, monthly, and quarterly terms
Service Model
What exactly does VMKeep provide?
Define the resource boundaries first, then decide whether it fits continuous builds, automation, or inference workloads.
Is this a Cloud Mac or a shared virtual machine?
VMKeep provides dedicated physical Cloud Macs. Each order maps to an independent physical node; its processor, memory, and local storage are not shared with other tenants. This is not a virtual machine. Use SSH for command-line tools and automation, or VNC to operate the macOS graphical interface.
Which workloads suit an always-available physical node?
Typical workloads include continuous iOS and macOS builds, self-hosted runners, React Native packaging, MLX model validation, remote graphical workstations, and automation that requires a fixed Apple Silicon environment. Nodes run continuously 365 days a year.
For short-term validation, rent by the day. If toolchains, caches, and model files need to remain available, monthly or quarterly planning is a better fit.
What resources are dedicated?
Dedicated means your order uses one specified physical Mac mini with its own chip, memory, and local storage. Network quality still depends on your access location, provider route, traffic, and target-service response, so assess remote interaction through real-world testing.
Model Catalog
Three configurations, with no models outside the catalog
The lineup includes two M4 configurations and one M4 Pro configuration. Compare memory, storage, and workload concurrency when choosing.
Which Mac mini configurations are available?
Only the following three configurations are currently available. 1TB SSD and 2TB SSD are storage add-ons, not separate fourth-tier models.
| Plan | Chip | Memory | Local storage | Best for |
|---|---|---|---|---|
| VMKeep M4 Core | M4 | 16GB | 256GB | Light builds, automation scripts, and initial quantized-model validation |
| VMKeep M4 Plus | M4 | 24GB | 512GB | Continuous CI, medium-sized models, and larger dependency caches |
| VMKeep M4 Pro | M4 Pro | 64GB | 2TB | High-memory MLX inference, parallel builds, and multi-task queues |
How should I choose between 16GB, 24GB, and 64GB?
- 16GB:Best for lightweight single-project builds, command-line automation, and small quantized-model validation.
- 24GB:Best for continuously running runners, medium-project dependency caches, and workflows that keep several development processes open.
- 64GB:Best for high-memory MLX inference, loading larger models, parallel workloads, or experiments requiring larger local datasets.
Model viability depends on more than parameter count. Estimate peak memory from quantization, context length, batch size, cache use, and other resident processes.
Available Locations
How to choose among five locations
Location affects the remote interaction route and the network path to external services, but not the physical specifications of the selected model.
Where does VMKeep currently offer nodes?
Available locations are Singapore, Japan (Tokyo), South Korea (Seoul), Hong Kong, and the US East, for a total of five nodes or facilities. All three models are listed in each location; actual availability is returned in real time by the console.
SingaporeSG · Southeast Asia
Japan (Tokyo)JP · East Asia
South Korea (Seoul)KR · Northeast Asia
Hong KongHK · South China & Southeast Asia
US EastUS-E · Eastern North America & Western Europe
Should the node be near the developer or the business service?
For SSH and VNC interaction, prioritize a location near the main operators with a stable network route. For automated builds, dependency downloads, or API calls, also consider the locations of your code repository, artifact storage, and target services. For teams working across regions, test the route by the day before committing to a longer term.
How can I confirm that a model and location are currently orderable?
Enter the order configuration flow, select a model, and check the locations returned in real time by the console. Do not use old screenshots or historical records saved elsewhere to determine current availability.
Rental Terms
How to choose daily, weekly, monthly, or quarterly billing
Match the term to workload duration, the value of reusing the environment, and your data-export plan—not just one price figure.
What is each billing term best for?
Daily
Short-term releases, compatibility testing, model feasibility checks, and temporary capacity increases.
Weekly
Focused iteration, migration rehearsals, milestone builds, and experiments lasting one to several weeks.
Monthly
Continuous CI, long-running runners, fixed development environments, and stable inference services.
Quarterly
Teams with stable workflows that continuously reuse the environment, caches, and data.
Which order details should I verify before renewal?
Verify the current model, location, rental term, storage add-ons, number of Thunderbolt 5 links, and order total. If the workload is finished, export build artifacts, model files, and required logs first, then remove keys, tokens, and signing materials.
Pricing
How much do the three models and add-ons cost?
The starting price is $19.2/day for VMKeep M4 Core. All prices below are in USD and are calculated according to the selected term.
What are the prices for the three Cloud Mac billing terms?
| Plan | Daily | Weekly | Monthly | Quarterly |
|---|---|---|---|---|
| VMKeep M4 Core | $19.2/day | $51.7/week | $95.8/month | $260.6/quarter |
| VMKeep M4 Plus | $39.4/day | $106.3/week | $196.9/month | $535.6/quarter |
| VMKeep M4 Pro | $60.8/day | $164.1/week | $303.8/month | $826.3/quarter |
How are expanded storage and Thunderbolt 5 links priced?
| Add-on | Daily | Weekly | Monthly | Quarterly |
|---|---|---|---|---|
| +1TB SSD | $2.5/day | $6.8/week | $12.6/month | $34.3/quarter |
| +2TB SSD | $5/day | $13.6/week | $25.2/month | $68.6/quarter |
| Thunderbolt 5 link (each) | $1.3/day | $3.5/week | $6.5/month | $17.7/quarter |
Add-ons are billed for the selected term. Before submitting, verify the quantity, term, and final total so temporary capacity is not carried into a long-term order unnecessarily.
USD Billing
Which payment methods are supported?
Only two payment categories are supported, and all orders are settled in US dollars (USD).
Which payment methods can I use?
We accept USDT-TRC20 and Visa / Mastercard / Amex (via Stripe) only. All payments are settled in US dollars (USD); the available gateway is determined by the console.
Which totals should I check before payment?
Verify the model, location, daily, weekly, monthly, or quarterly term, storage add-ons, number of Thunderbolt 5 links, and total in USD. For billing issues, submit a ticket in the console with the order number and relevant screenshots. Do not include passwords, private keys, or complete signing materials.
Remote Access
Which tasks suit SSH or VNC?
Command-line automation and graphical-interface work require different access methods. Confirm the node details and credentials before connecting for the first time.
Should I use SSH or VNC?
VNC
Best for configuring macOS graphical tools, checking interfaces, and remote development. Screen responsiveness depends on the network route and bandwidth between the operator and node.
What should I check before the first connection?
- Verify the node:Confirm that the location and host address in the console match the order.
- Verify the network:Check that your local network permits the required connection and rule out proxy, firewall, and provider-route issues.
- Update credentials:Change the password after first login and configure a separate key for SSH.
- Verify the base environment:Confirm disk space, system time, and Xcode and command-line tools versions.
- Save diagnostic details:Record the time, error output, reproduction steps, and node location for follow-up troubleshooting.
What troubleshooting sequence should I follow if connection fails?
First confirm the order and node details, then check your local network, address, and credentials. Next verify port connectivity, SSH key permissions, or VNC client settings as applicable. If the issue persists, retain sanitized command output, time, location, and reproduction steps and submit a console ticket.
AI & MLX
Run MLX inference services with confidence—model compatibility matters
The physical machine provides a fixed Apple Silicon environment, but compatibility, memory needs, and response performance must be tested with your model and parameters.
Can I deploy an MLX inference server on a Cloud Mac?
You can install the MLX environment, load a compatible model, start a listening service, and connect applications through an API. First validate model loading and a single request on the local loopback address, then configure the listening service, access controls, process supervision, and log archiving.
VMKeep makes no universal throughput or latency guarantees for any model. Actual performance depends on model architecture, quantization version, context length, output length, concurrency, batch size, memory pressure, and service implementation.
How should I choose memory for AI models?
Record the model file size, then measure resident memory after loading, peak generation memory, and cache growth with the target quantization. Start small validation with VMKeep M4 Core M4 / 16GB / 256GB; choose VMKeep M4 Plus M4 / 24GB / 512GB for ongoing services or medium models; evaluate VMKeep M4 Pro M4 Pro / 64GB / 2TB for high-memory inference and parallel workloads.
What should I check when an MLX service cannot be called?
- Confirm that the model path, configuration file, and quantization version are compatible with the current runtime.
- Check memory usage before and after loading to rule out peak memory exceeding the selected configuration.
- Send a request to the local loopback address first, then check the listening address, port, and access controls.
- Check process status, standard output, and error logs to confirm the service did not exit immediately after starting.
- Fix the input, output-length, and concurrency parameters, then record time to first token, sustained throughput, and peak resource usage.
How should I handle models, code, and build data?
Plan your own backups and encryption for repository access, model files, build artifacts, and business data. Apply least privilege to keys, tokens, and signing materials; never place them in public repositories or ordinary logs. Before the rental ends, export required data, revoke tokens, and remove keys.
For information about data handling on the website, in orders, or during support interactions, read the Privacy Policy. To report an issue, contact the service team through a console ticket or support@vmkeep.com.
Next Steps
Configurations and limits are clear—choose your model and location
Choose memory and storage for your workload, then verify live location availability and the order total in the console.