How to Run & Control Ollama (Local LLMs) on Mac from Your iPhone
One of the best reasons to own an Apple Silicon Mac is local AI. Thanks to unified memory on M-series chips, a basic Mac Mini or MacBook Pro can run models like Llama 3.3, Mistral, and DeepSeek Coder surprisingly fast without paying monthly API bills.
The default tool for running these models locally is Ollama. It gives you a clean CLI to download, manage, and chat with open weights right in your terminal.
The problem starts when you leave your desk. You might want to ask your local model a coding question from your phone, check how fast a prompt evaluates, or download a 14B parameter model while running an errand. Here is how to run and control your Mac's Ollama instance from your iPhone, and why the traditional setup advice is usually dangerous.
The Flawed Way: Exposing Port 11434
If you look up how to connect your phone to Ollama on Reddit or GitHub, most tutorials tell you to do this in Terminal:
launchctl setenv OLLAMA_HOST "0.0.0.0:11434"By default, Ollama only listens on 127.0.0.1 (localhost). Setting it to 0.0.0.0 tells Ollama to accept incoming connections from any device on your network.
Then the tutorials tell you to download a third-party iOS chat client, enter your Mac's local IP address, and port-forward port 11434 on your home router if you want it to work outside your house.
The Better Way: Direct Terminal Access via Macky
Instead of exposing an unauthenticated HTTP port or configuring complex reverse proxies, the cleanest way to interact with Ollama is to run it in your native Mac shell directly from your iPhone.
With Macky, your iPhone connects straight into your Mac's zsh shell over peer-to-peer WebRTC. All data is end-to-end encrypted with DTLS-SRTP, and you do not need to open any ports or alter Ollama's host environment variables.
How the Workflow Looks:
- Launch Macky on your iPhone and tap to connect to your Mac.
- Type your regular command, for example:
ollama run deepseek-coder:6.7b. - Type your prompt using your phone keyboard, complete with code snippets and file paths.
- Watch tokens stream back in real time with zero network lag.
Useful Ollama Commands to Run from Mobile
Having direct terminal access on your phone makes server maintenance effortless. Here are four common tasks you can trigger from anywhere:
Check which models are currently downloaded on your Mac and how much disk space they consume.
See which model is currently resident in your Mac's unified memory and how much VRAM it is occupying.
Kick off a 9 GB model download while you are away from home so it is ready to use when you sit down at your desk.
Immediately unload a large model from memory if you need to free up RAM on your Mac for other tasks.
What About Web UIs Like Open WebUI?
Another popular route is running a web interface like Open WebUI in a Docker container on your Mac. It provides a slick ChatGPT-style interface with chat history, system prompts, and document uploads.
If you choose that route, you still face the same networking dilemma: how do you access the web interface from outside your home?
You can run Tailscale on both devices to reach your local Docker container IP, or use Macky's full-screen desktop mode to view your browser directly. But for quick coding queries, model checks, and shell scripts, nothing beats a fast terminal session that opens in three seconds.
Get Started in Two Minutes
If you already have Ollama installed on your Mac, you are 90% of the way there. Install the Macky host on your Mac and the iOS app on your phone, and you can chat with your local models from anywhere in the world.