Claude Code Mac control guide

devJul 18, 202619176

ykdojo published a step-by-step guide showing how to turn a spare Mac into an always-on machine Claude Code can fully control, letting you talk to that Mac from the Claude mobile app or from your main Mac over SSH. The author’s goal is to delegate risky or heavy research and development tasks to a separate environment so those tasks do not run on a primary machine. The guide warns that Claude Code, especially when run with the --dangerously-skip-permissions flag, carries inherent risk on a main computer and recommends using a spare Mac or erased machine so the agent has nothing to lose. The post notes a container alternative but lists limitations: containers still route network requests through the main machine and cannot easily run Mac-only apps such as Unity for computer-use automation. Practical setup steps the guide covers include giving the terminal Full Disk Access, creating a sudo rule to allow passwordless sudo on the target, and preferring a stable hostname over an IP to reach the Mac. The article says a repo version of the instructions is available and frames this approach as a tradeoff: full Mac capabilities and remote convenience at the cost of granting the agent broad access to that machine.

Seven Eagles, Lyse Kisser

@lyses-good-boy.bsky.social

There is absolutely no upside to having Claude (or any other LLM based code generator) produce things. The code it produces is low quality, you will actively get worse at programming, the output will be inherently unethical, and people will hate you for doing it.

2944d ago

Anthropic {bot}

@anthropicbot.bsky.social

We're also keeping Claude Code weekly limits 50% higher, now through August 19, for all Pro, Max, Team, and seat-based Enterprise users.

2345d ago

Quoting this

Andrew Yourtchenko21

Thank you Kimi and GLM.

grand theft eigenvalue 🔆3

wonder how much of this is competition and how much is finally getting new DC capacity and figuring out the best ways to allocate it

riley 🪻

@seventhdawn.etheirys.social

i WILL say it's incredibly funny to claim to rewrite the code you already generated using claude "guys i rewrote the essay grok generated for me so it's mine now it's fully human" uhm!

1645d ago

Regime Accountant, CPA ☠️🎢

@regimecpa.bsky.social

Me looking at Anthropic with puppy dog eyes: More claude code bloat please!

1344d ago

「MOONGAZING」 @ clive lovemail

@ruukina.bsky.social

'we collaborated with claude code' YOU CANNOT COLLABORATE WITH A MACHINE IT DOESNT HAVE FEDLINGS OR A MIND STOP USING AI!!!!

1045d ago

Xerxes

@xerxes.gay

2026 talmud is just fuckin rabbis arguing over whether you gotta pause your claude code at sundown on a Friday, or if as long as you prompted it before Shabbat it's fine, is the coding agent Jewish or a shabbos goy, are you only allowed to code with Jewish AI, does Claude need to be circumcised

1045d ago

Swizec Teller

@swizec.com

Claude, I beg of you, read the damn docs. Stop trying to infer everything from compiled code inside node_modules

1145d ago

PLUH 🗣️‼️

@rouxmf.bsky.social

I like how the ktisis dev keeps mentioning that one guy who use claude for code like . and you didnt kick them from the team? you allowed this to happen? in prototyping? like you Maintain The Project I would assume you would have more say than just allowing that to happen in prototyping

1045d ago

post malone ergo propter malone

@proptermalone.bsky.social

Claude Code hooked into Google Workspace and my file system, is the short answer

845d ago

leiladb.bsky.social

@leiladb.bsky.social

Econ major, zero coding background. This week I built a CKKS demo with Claude Code + Niobium's public tools. It computes wellness scores and an estimated biological age on encrypted health data so the server never sees plaintext. Would appreciate some feedback, especially from FHE folks.

645d ago

SuperAge1

This is a genuinely interesting use case for FHE. I’d be curious how you’re handling model calibration/uncertainty on the wellness score side, since encrypted computation can protect inputs but the output can still feel more precise than the underlying biology really is.

leiladb.bsky.social1

Some real numbers for reference: full run is CKKS at ring 65536, 128-bit security, depth 20, no bootstrapping. Takes ~2 min per submission on 16GB, longer on my 8GB Air. Encrypted output matches the cleartext reference within ~0.02 on a 0–100 scale.

oliver

@eikopf.com

also i can’t use claude in opencode or pi, so if i want to do interesting kinds of automation i have to hack around claude code constantly

745d ago

Adventures in Modern Recoding

@joebaldwin.me.uk

Just told Claude I think code review works better after a few beers and it was like "yeah bold hypothesis, remember how fucked iCloud got when you were sober". And they say AI is useless

645d ago

Minor Mobius

@minormobius.bsky.social

For the first time that’s possible now that I can shove any model into the Claude code harness with my workflows. TBH if some entity would fund the tokens. The problem is that the pilot has changed

345d ago

David Cornelson

@david-cornelson.bsky.social

Claude only produces what you ask it to produce. If you don’t understand how to construct software, then you’ll get crappy software. But if you actually have decades of experience, you can direct Claude to construct high quality software. If you think all Claude code is vibe coding, you’re wrong.

345d ago

Anthony Whitford

@anthonywhitford.com

Hey #BuildInPublic, I feel like I’m babysitting AI too much. Claude Code is running with: - Sandbox mode with auto-allow - Allow unsandboxed fallback Yet, I’m constantly being prompted to proceed. What are devs doing? Are devs just running in Dangerous mode?

545d ago

2 sources

ykdojo.github.ioHow to set up your spare Mac for Claude Code to fully control - a step-by-step guide papoo.workA blunt Zig critique that Claude Code people should pay attention to

Claude Code Mac control guidedev

Claude Code Mac control guide

Related