Skip to content

THE FIELD JOURNAL

Ollama Android and iPhone Apps: Use Your Home PC's Models Anywhere

By Local AI Scout · Beginner guide

Hand-drawn phone showing a chat, linked to a chat window and a dark terminal window that stands for the home PC running Ollama, for a guide to using Ollama from an Android phone or iPhone

Can you use Ollama on Android or iPhone?

Yes, as a remote control. Ollama itself runs on macOS, Linux and Windows, and it serves an API on port 11434. An Android or iPhone app, or Open WebUI in the phone's browser, sends your prompts to that PC over home Wi-Fi, or over Tailscale when you are away. Running Ollama on the phone itself is unofficial.

This guide follows one reader: someone with a Windows 11 desktop and a 16GB graphics card that already runs Ollama, with gemma4:12b, a 7.7 to 8.0GB download, as the everyday model. They carry an iPhone, share the house with an Android phone, and want those answers on the sofa and on the train without opening the PC to the internet. A Mac mini owner gets notes along the way. New to Ollama? Start with our Ollama install guide for Windows, Mac and Linux. If you would rather run a small model on the phone itself, with no PC, see our guide to running AI on your phone offline.

Choose a direct Ollama app if:

Choose Open WebUI in the phone's browser if:

How does using Ollama from your phone work?

Ollama is a server program that waits for requests through its API (application programming interface), the web addresses apps use to send a prompt and get a reply. By default it binds 127.0.0.1 port 11434, the loopback address, so only programs on the PC itself can connect.

A phone app is just another client of that API: it asks for the list of local models, sends your message and shows the reply, while the graphics card does the work. So either Ollama listens on the network, or Open WebUI relays for it from the same PC, talking to Ollama on 127.0.0.1.

Diagram of Ollama on Android and iPhone: phones reach Open WebUI on port 8080 over home Wi-Fi or Tailscale, and Open WebUI talks to Ollama on 127.0.0.1 port 11434
The phone only sends text; the PC's graphics card runs the model, and Ollama never has to leave 127.0.0.1.

The route matters because local requests need no API key: anyone who can reach port 11434 can chat with your models, download new ones or delete them.

How do I connect my phone to Ollama on home Wi-Fi?

These steps set up the direct route for apps such as Reins or Maid, with the phone and PC on the same Wi-Fi. The PC's address here, 192.168.1.50, is an example.

  1. Let Ollama listen on the network. In Ollama's Settings, switch on Expose Ollama to the network, described in version 0.40.13 as allow other devices or services to access Ollama. It sets OLLAMA_HOST to 0.0.0.0, meaning every network interface on the PC. The manual route: on Windows, quit Ollama, add OLLAMA_HOST under Edit environment variables for your account, and start Ollama from the Start menu, with the value 0.0.0.0:11434 from the FAQ's examples; on a Mac, use launchctl setenv below.
  2. Open the firewall for port 11434. On Windows, run the rule below in PowerShell as administrator. It uses New-NetFirewallRule to allow the port only on networks set to Private, and only from your local subnet and from 100.64.0.0/10, the range Tailscale assigns by default. Windows 11 sets new networks to Public, so in Settings, Network & internet, open your home Wi-Fi or Ethernet connection and under Network profile type, select Private network, which Microsoft says is for networks where you know and trust the people and devices. Tailscale's Windows app marks its own network adapter as private, so the rule covers it too. On a Mac with the firewall on, open System Settings, Network, Firewall, then Options, and add Ollama to the apps allowed incoming connections.
  3. Find the PC's address. On Windows, run ipconfig in PowerShell, which displays the IPv4 address for each adapter, and use the Wi-Fi or Ethernet one; Settings, Network & internet, then your connection, also lists it next to IPv4 address. On a Mac on Wi-Fi, open System Settings, Wi-Fi, then Details next to your network, which shows the IP and router addresses; on Ethernet, look under System Settings, Network, your connection, Details, TCP/IP.
  4. Test from the phone's browser. Open http://192.168.1.50:11434. A reply reading Ollama is running proves the network path works before any app is involved.
  5. Add the server in your app. Enter http://192.168.1.50:11434 as the Ollama server address, then pick gemma4:12b from the model list. On an iPhone, tap Allow when the app asks to find devices on your local network, which iOS requires of any app that wants to interact with devices on your network.
# Windows, PowerShell as administrator: port 11434 on Private networks, from home network and Tailscale only
New-NetFirewallRule -DisplayName "Ollama 11434 (home network and Tailscale)" -Direction Inbound -Protocol TCP -LocalPort 11434 -RemoteAddress LocalSubnet,100.64.0.0/10 -Profile Private -Action Allow

# Mac, Terminal: the FAQ's alternative to the Settings switch, then restart Ollama
launchctl setenv OLLAMA_HOST "0.0.0.0:11434"

Is it safe to set OLLAMA_HOST to 0.0.0.0?

On a home network you trust, it is a fair trade, but every device on the Wi-Fi, including a visitor's laptop or a poorly patched smart TV, can use the API. I would do it only on a desktop that never leaves the house. On a laptop, switch Expose off before it joins cafe Wi-Fi, where your local subnet means strangers; leaving that network set to Public also keeps the Windows rule above inactive there, because a rule is active only when the specified profile is currently active.

The safer setup: Open WebUI with accounts

Open WebUI is a free, self-hosted chat interface that automatically attempts to connect to your Ollama instance; our Open WebUI install and update guide covers it in full. Installed with pip or uv, as our Gemma 4 guide does, open-webui serve runs on the same PC, at http://localhost:8080, and reaches Ollama on 127.0.0.1. Ollama's Expose switch can stay off.

The serve command listens on host 0.0.0.0 and port 8080 by default, so the phone can open http://192.168.1.50:8080 once the firewall allows port 8080 (repeat the rule above with 8080 in place of 11434), the local IP address and port pattern its docs describe. If you run it in Docker instead, its docs say Ollama on the host has to listen on 0.0.0.0, which loses this advantage.

Accounts are the real gain. The first user to register becomes the Administrator, which switches off new sign-ups until an admin turns them back on, and even then new accounts wait in a Pending state until an admin approves them. A pending user must be promoted to user or admin before they can chat, so a guest on your Wi-Fi sees a login page, not your models.

Install Open WebUI as an app on your phone

Every Open WebUI instance is a Progressive Web App (PWA): a website you install with its own icon, which opens full screen.

How do I use Ollama away from home without port forwarding?

Why not forward port 11434 on your router?

Port forwarding is a router rule that sends internet traffic on one port to a computer at home. For Ollama, that hands a password-free API to the whole internet, and people scan for it. A joint study by SentinelLABS and Censys, published in January 2026, recorded 175,108 unique Ollama hosts across 130 countries over 293 days of scanning. Fixed-access telecom networks, which include consumer internet providers, made up 56% of those hosts.

Diagram comparing a router forwarding Ollama port 11434 to anyone online with a phone reaching Ollama through a private Tailscale tailnet
Port forwarding publishes Ollama to every scanner online; Tailscale lets in only devices signed in to your own account.

The same goes for tunnels. Ollama's FAQ shows how to reach it with ngrok or Cloudflare Tunnel, and Enchanted's README suggests ngrok to make your Ollama server publicly accessible. As written, both give Ollama a public address with no password in front of it.

Set up Tailscale on the PC and the phone

Tailscale is a mesh VPN: it builds a private, encrypted network, called a tailnet, between devices signed in to your account, and gives each one a 100.x.y.z address that works even behind a firewall, with no router changes. The Personal plan is free for up to 6 users with unlimited user devices, for non-commercial use.

  1. Install Tailscale on the PC. Windows needs Windows 10 or later; run the .exe installer and log in from the tray icon. A Mac needs macOS 12 or later, and Tailscale recommends its Standalone variant.
  2. Install it on the phone. It works with iOS 15 or later and Android 8.0 or later. Tap Get Started, accept the VPN configuration, and log in with the same account.
  3. Find the PC's tailnet name. MagicDNS, on by default in tailnets created since October 20, 2022, lets your devices use the machine name instead of an address. tailscale status lists your devices, and tailscale ip -4 prints this device's 100.x.y.z address.
  4. Point the app at that name. For a direct app with Expose on, use http://my-pc:11434, replacing my-pc with your machine name. Leave Tailscale connected on the phone and it works on mobile data, and the address stays the same while the device remains registered.

Use Tailscale Serve for Open WebUI

Tailscale Serve routes traffic from other devices on your tailnet to a local service, over HTTPS with an automatic certificate. It needs HTTPS certificates enabled for your tailnet, and the command offers to turn them on. On the PC, run:

tailscale serve --bg 8080
tailscale serve status

The serve command accepts a plain port number, and with --bg it keeps running and resumes after a reboot. Your phone then opens an address like https://my-pc.tail1234.ts.net at home or away. Open WebUI's own Tailscale guide notes that browser features such as voice calls require HTTPS, and suggests setting WEBUI_URL and CORS_ALLOW_ORIGIN to that address.

Two cautions. Turning on HTTPS means machine names are published in the public Certificate Transparency ledger, so rename a PC called after a person first. And to keep Open WebUI off the home network entirely, start it with --host 127.0.0.1, an option the quick start describes for changing the bind address; Serve still reaches it, because it proxies to http://127.0.0.1.

Info card of Ollama and Tailscale commands: launchctl setenv OLLAMA_HOST, a Windows firewall rule for port 11434, ipconfig, tailscale ip -4 and tailscale serve --bg 8080
Expose Ollama only to your home network and tailnet, test the address from the phone, and never forward the port.

Why not Tailscale Funnel?

Funnel is Serve's public sibling: it routes traffic from the broader internet to a local service, for anyone, even if they don't use Tailscale. Open WebUI's docs warn that Funnel makes your Open WebUI accessible to anyone on the internet. Pointed at Ollama, it recreates the port-forwarding problem.

Which Ollama app should you use on Android and iPhone?

Ollama's README lists phone clients from other developers under Community Integrations. Here is where they stood in October 2026.

AppPhonesPriceOpen sourceNotable featuresLast update
ReinsiPhone and iPad (also Mac, Vision Pro)Free; optional Pro at $6.99 a month, $39.99 a year or $99.99 onceOnly up to version 1.3.4, under GPL-3.0Several servers with API keys, web search tool, file attachments, on-device MLX models2.3.3, October 21, 2026
Enchanted Developers OnlyiPhone, iPad, Mac, Vision ProFreeYes, Apache 2.0Image attachments, voice prompts, read aloud1.9.0, August 19, 2025
MaidAndroidFreeYes, MITOllama plus cloud APIs; also runs GGUF models on the phoneGoogle Play, September 2, 2026
ChatterUIAndroid (APK from GitHub)FreeYes, AGPL-3.0Remote mode with an Ollama template, character cards, on-device GGUFv0.10.0, September 22, 2026
Ollama AppAndroid (APK), plus Windows and LinuxFreeYes, Apache 2.0Simple client for Ollama only1.2.0, September 14, 2024
Open WebUIAny phone browser, installableFreeSource available under the Open WebUI LicenseAccounts, chat history on the PC, documents, voicev0.12.4, October 25, 2026

Dates come from the store listings and the ChatterUI, Ollama App and Open WebUI release pages; Ollama App 1.2.0 is also the latest build on IzzyOnDroid.

Card comparing Ollama Android and iPhone apps in October 2026: Reins and Enchanted for iPhone, Maid, ChatterUI and Ollama App for Android, Open WebUI for both
Reins and Maid are the maintained native picks; Open WebUI works on any phone with a browser.

Is there an official Ollama app for Android or iOS?

No. The download page offers macOS, Linux and Windows. Every phone app here comes from another developer, so check its privacy label and update date before you trust it with your chats.

The best Ollama client for iPhone

Reins is the one I would install. It is actively updated, its App Store label says the developer does not collect any data, and its privacy policy says prompts go directly to the server address you provide. Its terms describe optional Reins Pro purchases for extra features. Enchanted was renamed Enchanted Developers Only, has not changed since August 2025, and its README now names Jaz as the new iteration, which is an app for running coding agents, not an Ollama chat client.

The best Ollama app for Android

Maid is the easiest, because it comes from Google Play, where its data safety section says no data collected. ChatterUI is more configurable, but you install its APK by hand from GitHub, and its README says iOS is currently unavailable. Ollama App's README explains that it does not host an Ollama server on device, which is right for this job, but with no release since 2024 I would treat it as a fallback.

Can Ollama run on Android or iPhone itself?

Not officially. Ollama's docs cover Linux, macOS, Windows and Docker and nothing for phones. On Android, the unofficial route is Termux, a Linux terminal app: its package repository carries ollama 0.40.11, maintained by the Termux team, not by Ollama. Termux itself warns that it may be unstable on Android 12+, because Android kills extra background processes and any that use excessive CPU.

Ollama's README also lists Ollama Android Chat, which starts Ollama on Android without Termux; its last tagged version, v0.0.5, dates from May 2025. On iPhone there is no Ollama build. Reins can run MLX models on the iPhone and share them over an Ollama-compatible endpoint, but its policy warns that any device on your local network can connect without a password.

On-device models stay small, because a phone has far less memory than our reader's graphics card. For AI with no PC at all, start with our hub on running AI on your phone offline and the guides to local LLM apps on Android and local LLM apps on iPhone. For an always-on machine serving the whole house, see our home AI server guide.

Why can't my phone connect to Ollama?

Connection refused

Ollama is still on 127.0.0.1. Check that Expose is on, or OLLAMA_HOST is set, then quit and restart Ollama. First confirm that http://localhost:11434 shows Ollama is running in a browser on the PC itself.

It worked yesterday: wrong IP address

Windows recommends DHCP, which automatically assigns IP addresses to the devices on your network, so the PC's address can change over time. Run ipconfig again and update the app, or use the Tailscale name, which does not change.

The page times out: firewall

If localhost works on the PC but the phone waits and gives up, the firewall is dropping the request. Check that the rule for 11434, or 8080 for Open WebUI, exists, and that your home network is set to Private. Windows 11 sets new networks to Public by default, where your PC will be hidden from other devices on the network, and the rule above only applies to the Private profile.

Works on Wi-Fi, fails on mobile data

A home address such as 192.168.1.50 only works inside your home network. On mobile data, switch the app to the Tailscale name or the Open WebUI https address, and check that the Tailscale app on the phone shows it is connected.

The app cannot see the PC at all

On an iPhone, open Settings, Privacy & Security, Local Network and turn access on for the app. On Android 17, apps that target it need a local network permission in the Nearby devices group, so allow Nearby devices if the app asks.

403 Forbidden through Tailscale Serve or a proxy

While Ollama listens only on 127.0.0.1, its server code rejects requests whose host name is not a local one, and Serve's proxy code passes the original host name through, so pointing tailscale serve at port 11434 is likely to return 403. The FAQ's proxy examples, for the same situation, set the Host header to localhost:11434 when Ollama is not exposed on the network. Use Serve for Open WebUI, and http://my-pc:11434 with Expose on for direct apps.

Common questions about Ollama on Android and iPhone

Can Ollama run on Android?

Not officially. Ollama's docs cover Linux, macOS, Windows and Docker only. The Termux terminal app packages Ollama unofficially, and Termux warns it may be unstable on Android 12 and later. The practical approach is Ollama on a PC with an Android app such as Maid as the client.

Is there an official Ollama app for iPhone?

No. Ollama offers downloads for macOS, Linux and Windows only. iPhone apps such as Reins and Enchanted come from other developers and connect to an Ollama server you run on a computer. Reins is the actively updated one; Enchanted has not been updated since August 2025.

How do I connect my phone to Ollama on my PC?

Turn on Expose Ollama to the network in Ollama's settings, or set OLLAMA_HOST to 0.0.0.0, and restart Ollama. Allow port 11434 in the firewall for your Private home network, find the PC's IP address with ipconfig, then enter that address with port 11434 as the server in your phone app.

Is it safe to expose Ollama with OLLAMA_HOST set to 0.0.0.0?

Only on a home network you trust. Ollama's local API does not require authentication, so any device that can reach port 11434 can chat with, download or delete your models. Never forward that port on your router. Open WebUI with accounts is safer.

How do I use Ollama away from home?

Install Tailscale on the PC and the phone and sign in with the same account. The phone can then reach the PC by its Tailscale machine name over mobile data, with no router changes. For Open WebUI, tailscale serve adds an HTTPS address only your devices can open.

What is the best Ollama app for Android?

Maid is the simplest choice, because it is free and open source, installs from Google Play and connects to Ollama as a remote provider. ChatterUI offers more settings but installs as an APK from GitHub. Open WebUI in Chrome, installed to the home screen, also works well.

My verdict for the Windows desktop and its phones: Open WebUI over Tailscale

For our reader, the best setup keeps Ollama where it starts, on 127.0.0.1, with Open WebUI in front of it and an account for each person. Tailscale Serve then gives both phones one HTTPS address, installed to the home screen, that works on the sofa and on the train. Nothing is forwarded on the router, and a guest on the Wi-Fi meets a login page.

If the phone is only yours and you want a native app, switch on Expose, add the firewall rule, set your home network to Private, and use Reins on the iPhone or Maid on Android, with the Tailscale name for trips away. The Mac mini works the same way. Never forward port 11434 to the internet.

Your next step: install Tailscale on the desktop and your phone, sign in to both with the same account, run tailscale serve --bg 8080, then turn off Wi-Fi on the phone and open the https address it prints.

Sources

Keep exploring: Apps · Models · All guides

THE SCOUT LETTER

Good finds.
Less noise.

Useful local AI apps, practical guides and the updates worth your time. A thoughtful note in your inbox.

Free to join. Unsubscribe with one click.