Gemini CLI on iPad: the 2 Ways That Work

You can run Gemini CLI on an iPad, but not the way you'd guess. iPadOS won't run it locally, and unlike Android there's no Termux to fall back on. The fastest path is Cosyra on the App Store. The same iOS app runs natively on iPad Pro, iPad Air, and iPad mini, and the Ubuntu container it provisions has Google's gemini CLI already installed. Pair a Magic Keyboard, paste a paid Gemini API key, type gemini, and the agent is up. Sign up gets you 1 hour free, no credit card. The only other path that works is SSH into your own always-on machine, and the reason the obvious on-device options fail is worth understanding before you waste an afternoon on them.

This guide is the iPad-specific companion to our Gemini CLI on your phone walkthrough. The phone version covers iPhone and Android generically, and the Gemini CLI on Android guide goes deep on the Termux-plus-NDK path that only Android supports. This one focuses on what's different about the iPad: it looks the most like a laptop, yet it's the hardest device to run a local terminal agent on, because the one on-device tool that makes Android work, Termux, does not exist here. If you want the broader picture of all four AI agents on mobile, start there. Our Claude Code on iPad guide is the sibling to this one for the same form factor, Gemini CLI on iPhone covers the same agent on the smaller one-handed screen, and Claude Code vs Gemini CLI lines the two agents up if you're still choosing.

Why is "Gemini CLI on iPad" different from Android?

Gemini CLI on iPad is a different problem from Android because of one missing tool: Termux. On Android, Termux gives you an on-device package manager and a real (if fiddly) Linux userland, so you can install Node and fight the agent onto the phone itself. iPadOS has no equivalent. Apple's sandbox doesn't allow a general-purpose package manager, so the two terminal apps people reach for — iSH and a-Shell — work very differently, and neither runs Gemini CLI. That's not a Cosyra opinion; it's two open GitHub issues, which we walk through below.

The second thing that changed is the price of entry. Until 2026-06-18, you could sign in to Gemini CLI with a personal Google account and get a generous free quota with no API key. Google retired that path that day — the no-key login now returns HTTP 410 Gone, and the official successor is Antigravity CLI. Per Google's own transition announcement, the open-source gemini binary keeps working, but only with a paid key from Google AI Studio or Vertex AI. So every iPad path below assumes you'll paste a paid GEMINI_API_KEY. The binary itself is alive and shipping fast: the latest release was v0.49.0 on 2026-06-25, with 105,605 GitHub stars as of 2026-06-28. Google just took away the free first-party endpoint, not the toolchain.

Why won't Gemini CLI run on iSH or a-Shell?

Gemini CLI won't run on iSH or a-Shell because both lack a working modern Node.js runtime, and Gemini CLI is a Node.js binary. The failure modes are specific and documented, not vague "compatibility issues."

• iSH crashes Node with "Illegal instruction." iSH runs Alpine Linux through an i386 user-mode emulator on ARM hardware. Modern Node.js prebuilt binaries emit CPU instructions the emulator doesn't implement, so node --version dies with Illegal instruction. This is tracked in iSH issue #2335, open since January 2024. A second issue, #2604, is literally titled "Gemini cli" — a user asking how to install it, with a reply that current Node/npm versions aren't supported and no resolution since. We tried it on an iPad to confirm; node crashes before Gemini CLI ever gets a chance to load.
• a-Shell ships no Node.js and forbids fork. a-Shell bundles native iOS-compiled utilities, which makes it fast, but it doesn't include a Node.js runtime, and iOS sandboxing forbids fork(2) — the subprocess model agents rely on to run shell commands. There's no apt or apk to add Node either. It's a genuinely useful Unix scratchpad; it is not a host for an AI coding agent.

The honest takeaway: if you bought an iPad expecting it to be a smaller laptop for terminal work, the on-device story is worse than on a $200 Android phone, where Termux at least gives you a fighting chance. The fix isn't a cleverer install command — it's running the CLI somewhere that has a real Linux kernel. If you want the full field of iPad coding setups beyond Gemini, our coding on iPad guide lines every option up.

What are the two real ways to run Gemini CLI on iPad?

There are two real ways to run Gemini CLI from an iPad as of 2026-06-28: a cloud container with Gemini CLI pre-installed (Cosyra), or SSH from a terminal app into your own always-on machine. They're not interchangeable — one needs a machine of your own, one doesn't. We walk through both.

1. Cosyra (cloud terminal, Gemini CLI pre-installed)

This is what we build. The Cosyra iOS app is a universal binary that runs on iPad with the full iPadOS UI — multitasking, Slide Over, Split View, Stage Manager, external display. You get a persistent Ubuntu 24.04 x86_64 container with Gemini CLI, Claude Code, Codex CLI, and OpenCode already installed, 30 GB storage, and session hibernation that resumes exactly where you left off. Because the container is real glibc x86_64 with Node 20+, the emulation and NDK problems that kill the on-device shells simply don't exist. The same container is reachable from iPhone, Android, or web, so you can start a Gemini session on the iPad and pick it up on a phone in a waiting room.

• Works when: you want zero setup on iPad and Gemini CLI running on a real Linux host without owning one.
• Breaks when: you have no internet (the container is in the cloud), or you need a local macOS toolchain like Xcode (the container is Ubuntu, not macOS).
• Cost: 1 hour free on signup, no credit card. Extend with a 10-hour, 7-day trial when you want more. After that, $29.99/month or $300/year. You bring your own paid Gemini key and Google bills you for tokens. See pricing.

2. SSH from a terminal app to your own machine

The classic remote-into-your-own-box path. Blink Shell is a paid iOS terminal with Mosh and tmux support refined over years. You SSH into a desktop, Mac, or VPS that has Gemini CLI installed, run tmux so the session survives app switching, paste your paid key on the host, and drive gemini from there. This is a fine setup if you already keep a Linux box awake — but you're now maintaining the box, the key, and the tunnel, and the moment the host sleeps the iPad has nothing to talk to.

• Works when: you already have an always-on Linux box or Mac and are comfortable with SSH plus tmux.
• Breaks when: the host sleeps or the network drops. Chaining the iPad to a machine in the closet defeats the portability that made you reach for an iPad on the train in the first place.
• Cost: the terminal app (Blink is a paid subscription) plus whatever the host machine costs to run.

How do you set up Gemini CLI on iPad with Cosyra?

You set up Gemini CLI on iPad in about three minutes: install Cosyra from the App Store, pair a Magic Keyboard, paste a paid Gemini API key, and type gemini. No npm install, no emulator, no SSH tunnel, no Mac. The steps below are exactly what we run on a fresh iPad.

Step 1: Install Cosyra on iPad

Open the App Store on your iPad and search for Cosyra. The app is a universal iOS binary, so the same listing covers iPhone, iPad Pro, iPad Air, and iPad mini. Sign in with Apple, Google, or email and the app provisions a fresh Ubuntu 24.04 x86_64 container on first launch, with Gemini CLI already on the PATH. The first launch takes about 15 seconds in our testing.

Step 2: Pair a Magic Keyboard and arrange Split View

Attach the Magic Keyboard or any Bluetooth keyboard with a trackpad. The trackpad is the part that matters: it gives you a cursor for text selection, link tapping, and scrollback that the on-screen interface can't. Then pull a second app in as Split View — we usually keep Safari on the right with the repo or AI Studio open, and Cosyra on the left with the terminal. On M-series iPads (any M1/M2/M3/M4 iPad Pro or iPad Air from 2022 onward) you can switch to Stage Manager for floating, resizable windows and a second display over USB-C.

Step 3: Add a paid Gemini API key and run gemini

Since the free personal-account login retired on 2026-06-18, Gemini CLI now needs a paid key. Generate one at Google AI Studio (or use a Vertex AI key), then in the Cosyra terminal:

Clone a repo and start a session:

Gemini reads the repo and starts working. On iPad, the larger screen means you can read the full diff without scrolling, which is the single biggest ergonomic win over the phone. One Gemini-specific note: with Gemini 2.5 Pro selected on a paid key, you keep the 1-million-token context window, which is genuinely useful when you point the agent at a large codebase and ask it to trace something across many files.

Which iPad keyboard shortcuts matter for Gemini CLI?

The iPad shortcuts that matter for Gemini CLI are the iPadOS multitasking ones, not in-app ones. Cosyra is a normal terminal — Ctrl-C, Ctrl-D, and Cmd-K (clear) all work once a hardware keyboard is attached. The system-level ones worth memorizing:

• Globe key + Q: show the App Library to pull Safari in next to Cosyra.
• Globe key + Up arrow: open the multitasking switcher to set up Split View or Stage Manager groups.
• Cmd + Tab: cycle apps, the same as macOS.
• Cmd + Space: Spotlight. Type "cosyra" to jump straight to the terminal from any app.
• Press and hold the mic key on the Magic Keyboard to dictate a task. Dictation is faster than typing for the natural-language part of an agent session, and Gemini tasks are usually a sentence or two.

What does iPad Gemini CLI look like in practice?

iPad Gemini CLI in practice looks like Cosyra on the left with the terminal, Safari on the right with the repo or docs, and short typed or spoken tasks driving the agent. Two workflows we run from an iPad regularly:

Couch feature work with Split View

Evening on the couch, you want a small endpoint added before tomorrow. Open Cosyra, clone or pull the branch, drag Safari in with the repo, and hand Gemini the task:

Gemini writes the handler, wires the route, adds the test, and runs the suite. You read the result on the big screen and commit with git commit from the same terminal. No laptop in the loop, and nothing crashed on an emulator first.

Large-context refactor on an external display

Plug an M-series iPad Pro into a USB-C display and enable Stage Manager. Point Gemini at a big repo and lean on the 1M-token context: ask it to trace where a config value flows across modules before you change it. Run Gemini in one Cosyra window and Claude Code in a second on the tests, with Safari showing the staging environment. Because both agents live in the same Ubuntu container, they see the same files and git history — you're not syncing anything between them. The iPad is the workstation; the container is the source of truth.

How do the iPad Gemini CLI options compare?

The options compare cleanly when you line them up against what actually matters: does Gemini CLI run at all, does it need a machine of your own, and what happens when that machine sleeps. The table covers the two working paths plus the two on-device shells people try first and why they fail, as of 2026-06-28.

Choose Cosyra if you want Gemini CLI genuinely running with no machine of your own and the same container across iPad, iPhone, and Android. Choose Blink plus SSH if you already keep an always-on Linux box and want full control of your own host. iSH and a-Shell are in the table only so you don't lose an evening discovering what their issue trackers already say.

Frequently asked questions