How to use Ollama from your phone

You already pull models with Ollama on a Mac, Windows PC, or Linux box. The next want is obvious: chat with those same models from your phone — couch, another room, eventually away from home — without pasting prompts into a cloud chat account.

What "Ollama on your phone" actually means

Ollama does not ship a first-party iOS/Android chat app that replaces your desktop install. In practice you pick one of these:

• LAN client → home Ollama — model on your computer at port 11434. Fastest and private on the same Wi‑Fi.
• VPN / Tailscale / tunnel + same client — same API, reachable when you leave the house, without opening the port to the whole internet.
• Encrypted companion / QR relay — polished remote path (in LM Mini: Home Share with phone on Mac, Connect on Windows/Linux).
• On-device GGUF/MLX in a phone app — model on the phone itself. Related, but not "your Ollama."

Sibling backends show up the same way: LM Studio often serves near :1234, Jan AI near :1337. Many phone clients can point at more than one.

What you need

• A computer with Ollama and at least one model pulled
• Phone and computer on the same Wi‑Fi for the first test (guest networks and AP isolation often block device-to-device traffic)
• A mobile client that can call Ollama's HTTP API
• Willingness to set OLLAMA_HOST so Ollama listens on the LAN — not only on 127.0.0.1

Ports to remember:

• Ollama default: 11434 → http://YOUR-LAN-IP:11434
• LM Studio sibling: commonly 1234
• Jan AI sibling: commonly 1337
• On the phone, never use localhost — that is the phone itself, not your PC

Path A — Same Wi‑Fi (do this first)

Prove typed chat works on LAN before you touch VPN, tunnels, or voice features.

1. Make Ollama listen on the LAN

By default Ollama binds to loopback. Other devices cannot connect until you change the bind address with OLLAMA_HOST.

macOS (app install) — set the host, then fully quit and reopen Ollama from the menu bar:

launchctl setenv OLLAMA_HOST "0.0.0.0:11434"

Linux (systemd) — add Environment="OLLAMA_HOST=0.0.0.0:11434" under [Service] via systemctl edit ollama.service, then reload and restart the service.

Windows — set a user environment variable OLLAMA_HOST to 0.0.0.0:11434, then fully quit Ollama from the tray and start it again.

2. Find your computer's LAN IP

On the computer: macOS/Linux ipconfig getifaddr en0 or hostname -I; Windows ipconfig and look for the Wi‑Fi IPv4. Example: 192.168.1.42.

3. Point the phone at that URL

In your mobile client, add a server like http://192.168.1.42:11434 (your IP). Pick a pulled model. Send a short message.

If it fails: confirm Ollama is running, both devices share Wi‑Fi (not guest/AP isolation), the host bind was restarted, and the phone is not using localhost.

Path B — Away from home (honest options)

Do not port-forward :11434 to the public internet unless you really know what you are doing. Prefer:

1. Tailscale / WireGuard / similar — phone joins the same private mesh as your PC, then use the same LAN-style URL over the mesh IP.
2. SSH tunnel — power-user path; works, more fiddly on mobile.
3. LM Mini remote — on Mac, LM Mini Home → Share with phone; on Windows/Linux, LM Mini Connect. Encrypted relay with a key you control. Remote is a Pro feature because it costs me to run — help me out here.

Same Wi‑Fi never required a companion app. Away-from-home is where people get stuck, and where a dedicated path earns its keep.

Phone clients (pick what fits)

Several apps can talk to an Ollama-compatible API. Features differ: on-device models, multi-backend (Ollama + LM Studio + Jan), voice, group chat, tools/MCP, widgets.

LM Mini is one option if you want Ollama and LM Studio / Jan / oMLX / on-device in one app, plus optional remote. It is not the only client — Locally and others exist. Use what matches your stack.

Setup sketch in LM Mini: Settings → Server → add http://YOUR-LAN-IP:11434 → pick a loaded model. Away from home, pair Remote Access instead of typing a LAN IP. Details: Getting started and the docs hub.

FAQ

Does Ollama need internet to chat?

Not once the model is downloaded. Downloads need the net. Phones away from home need a path to your PC, not a public Ollama port. Longer answer: Does LM Studio need internet?

Can I run the Ollama binary itself on iPhone?

Not as a first-party Ollama app. On-device inference in phone apps uses other runtimes (GGUF/MLX). Your desk Ollama stays the host for "Ollama on phone" via API.

Ollama vs LM Studio vs Jan for phone?

Same pattern: server on PC, client on phone. Ports and UIs differ. LM Mini can point at more than one if you hop between them.

The bottom line

Start on Wi‑Fi. Fix OLLAMA_HOST. Use the LAN IP on the phone. Add VPN or LM Mini remote only when you leave the house. Keep :11434 off the public internet.

If you want the path I maintain day to day: download LM Mini, connect Ollama on LAN, then try Home / Connect when you travel.