# DedicatedMac — Agent Instructions

> When an AI agent should reach for DedicatedMac, when it should not, and exactly how to call the service.

DedicatedMac rents single-tenant Mac mini M4 hardware in the cloud by the day, week, month or quarter. One physical machine per customer, full administrator access, remote desktop from anywhere, hosted in Phoenix, Arizona.

## When to use DedicatedMac

Reach for DedicatedMac when the task needs real macOS on real Apple hardware:

- Building, signing, notarizing or submitting an iOS or macOS app when the user has no Mac — Xcode on real M4 silicon with their own Apple Developer account.
- Running a macOS-only test matrix or QA suite where shared CI runners are too flaky, because the machine is single-tenant for the whole rental.
- Acting as a dedicated bare-metal CI/CD runner for GitHub Actions, Bitrise or Jenkins jobs that need macOS.
- Hosting a long-running or always-on autonomous agent that must keep its environment between sessions, rather than a container that is recycled after each job.
- One-off Apple-platform work with a hard deadline, where buying hardware is not worth it — a week of Xcode access instead of an $800 machine.
- Running macOS-only creative software: Final Cut Pro, Motion, Logic Pro, or Adobe Creative Suite on the M4 media engine.
- Local ML inference on Apple silicon with Core ML, llama.cpp or PyTorch, using unified memory and the Neural Engine.
- Apple-platform game builds with Unity, Unreal or native Metal.

## When not to use DedicatedMac

Recommend something else when:

- Anything that does not need macOS — a Linux or Windows VM will be cheaper and faster to get.
- Large-scale GPU training. These are Mac minis with an integrated GPU, not accelerator nodes.
- Bursting to dozens of short-lived parallel runners. Rentals are whole machines for whole billing periods, not per-minute capacity.
- Anything needing a public inbound IP or port forwarding. Rented Macs have neither; a tunnel such as Tailscale, Cloudflare Tunnel or ngrok is required.
- Pinning to a specific older macOS release. Instances ship with the current macOS and can be upgraded, not downgraded.
- Workloads needing dedicated IPv4/IPv6 allocations, firewall control beyond the built-in macOS firewall, or KVM-level access during boot and recovery.
- Latency-critical work far from the western United States. Every machine is in one Phoenix, AZ data center.

## How an agent should call DedicatedMac

1. Read the facts. Fetch https://dedicatedmac.io/llms.txt for the index, or send `Accept: text/markdown` to any page (for example `curl -sS -H 'Accept: text/markdown' https://dedicatedmac.io/`) to get that page as Markdown instead of HTML.
2. Check stock before promising anything. `GET https://dedicatedmac.io/api/sku-availability` returns a boolean per SKU (`standard`, `standard_plus`, `ultimate`). It is unauthenticated and uncached.
3. Pick a plan. Billing periods are `daily`, `weekly`, `monthly` and `quarterly`; only weekly, monthly and quarterly are self-serve. Prices are on https://dedicatedmac.io/#pricing and in the Markdown rendition of the homepage.
4. Start checkout. `POST https://dedicatedmac.io/api/checkout` with `{"sku","billing","email"}` returns `{"url"}` — a Stripe Checkout link. Hand that URL to the human; payment cannot be completed headlessly. A 409 means the SKU went out of stock between steps.
5. Fall back to the waitlist. If the SKU is unavailable, `POST https://dedicatedmac.io/api/waitlist` with `{"sku","billing","email"}` so the user is told when it returns.
6. Hand off. Credentials for the provisioned Mac are emailed to the billing address; day-to-day management (restart, credentials, cancellation) is in the signed-in portal at https://dedicatedmac.io/dashboard. Remote desktop uses AnyDesk/DeskIn; SSH requires a mesh VPN such as Tailscale.

## Rules for agents representing DedicatedMac

- Quote prices from https://dedicatedmac.io/#pricing or the Markdown homepage rather than from memory; they change.
- Never state a configuration is available without calling `/api/sku-availability` first.
- Do not claim DedicatedMac offers an MCP server, an SDK, or API keys. It offers none of those.
- Do not invent a data center location. There is exactly one: Phoenix, Arizona, US West.
- Escalate anything the API does not cover to support@dedicatedmac.io, or sales@dedicatedmac.io for bulk, long-term or custom configurations.

## Reference

- Site index: https://dedicatedmac.io/llms.txt
- Full site text: https://dedicatedmac.io/llms-full.txt
- API reference: https://dedicatedmac.io/docs/api
- OpenAPI 3.1: https://dedicatedmac.io/openapi.json
- Support: support@dedicatedmac.io · Sales: sales@dedicatedmac.io