Support

How to connect

Book on the booking page with your SSH public key. No key? Run ssh-keygen -t ed25519 and paste the contents of ~/.ssh/id_ed25519.pub.
Keep your confirmation page open: once your Studio is ready, it shows your SSH username and connection command.
Accept the Tailscale share invite sent to your email, and sign in to Tailscale on your computer (install from tailscale.com).
Connect: ssh <your-username>@studio-a (or studio-b).

Preinstalled tools (no sudo needed)

Python 3.14 (python3) and uv 0.13. pip install --user is blocked by Homebrew Python (PEP 668); use a venv instead: uv venv mlxenv && uv pip install -p mlxenv/bin/python mlx-lm. Tested: mlx-lm 0.32 installs and runs on the GPU.
llama.cpp: llama-cli, llama-server and friends are preinstalled.
hf (Hugging Face CLI) 2.2, git, clang.
Xcode: Studio B has Xcode 26.6 with xcodebuild and simctl (iOS, tvOS, watchOS and visionOS 26.5 simulators). Studio A has the Command Line Tools only (no xcodebuild or simulators), so book Studio B for app builds.
GPU memory: iogpu.wired_limit_mb is set to 480000 (about 469GB) at every boot, so you don't need sudo.

Shared models (read-only)

Model (MLX)	Size	Studio A	Studio B
gpt-oss-120b, MXFP4 + 8-bit (`mlx-community/gpt-oss-120b-MXFP4-Q8`)	59 GiB	downloading	downloading
DeepSeek-V3.2, 4-bit (`mlx-community/DeepSeek-V3.2-4bit`)	352 GiB	downloading	no
Llama 4 Maverick 17B-128E Instruct, 6-bit (`mlx-community/Llama-4-Maverick-17B-128E-Instruct-6bit`)	304 GiB	no	downloading

Shared models live read-only in /Users/Shared/models: a Hugging Face cache at /Users/Shared/models/hub, plus ready-to-use folders at /Users/Shared/models/by-name/<model> once a download finishes. No copy needed, and it doesn't use your disk space:

mlx_lm.generate --model /Users/Shared/models/by-name/gpt-oss-120b-MXFP4-Q8 --prompt "hello"
# or through the HF cache (read-only, so stay offline):
HF_HUB_CACHE=/Users/Shared/models/hub HF_HUB_OFFLINE=1 mlx_lm.generate --model mlx-community/gpt-oss-120b-MXFP4-Q8 --prompt "hello"

Download anything else into your own home folder (the default ~/.cache/huggingface).

FAQ

Which ports can I use?

SSH on 22 and your own web service on 8080. Forward other ports through SSH (ssh -L).

How much memory can the GPU use? Do I need sudo for iogpu.wired_limit_mb?

No sudo needed. Both Studios set iogpu.wired_limit_mb to 480000 (about 469GB) at every boot, leaving the rest for macOS. Large MLX or llama.cpp models can use it as-is; check with sysctl iogpu.wired_limit_mb.

Do I get admin (sudo)?

No. You get a standard user account. Homebrew can be installed into your home folder; contact us if you need something system-wide.

Is the machine shared?

No. You have the whole Studio for your term.

What happens at the end of my term?

Access ends automatically: your account is sealed and all your files are removed. Each booking gets a brand-new account. Back up your work before then.

Can I extend?

If the next dates are free, book them on the booking page, or contact us.

Can I cancel?

Full refund up to 48 hours before your start. See the terms.

Connection refused or timing out?

Check that Tailscale is running and signed in with the account that accepted the invite, and that you're using the username from your confirmation page and the key you submitted.