Why I built a keyboard for the CLI renaissance

More people use the terminal now than at any point in the last decade, and most of them aren’t sysadmins.

Claude Code, OpenClaw, Gemini CLI, Codex — these tools live in a shell. If you want to use them, you need a terminal. And a growing number of people are accessing that terminal from their phones, SSHing into a Raspberry Pi or a Mac Mini from the couch, a waiting room, or the train.

The phone keyboards they’re using weren’t built for this.

What’s out there

The existing options fall into two categories, and none of them were designed for AI CLI workflows.

Terminal keyboards like Hacker’s Keyboard, Unexpected Keyboard, and BeHe Keyboard give you Ctrl, Alt, Esc, and arrow keys as a system keyboard. They’re built for traditional terminal power users — vim, emacs, sysadmin work. They handle key events well but none of them have voice input, slash command shortcuts, or image upload. Hacker’s Keyboard hasn’t been updated in years. Unexpected Keyboard is actively maintained and well-designed for what it does, but it’s targeting a different workflow.

SSH clients like Termius, ConnectBot, and JuiceSSH include their own terminal key toolbars. They work well inside their own apps, but the keys only exist inside those apps. If you use Cockpit in a browser, or Termux, or any other terminal tool, those toolbars don’t help you. Termux has its own configurable extra keys row, which is useful if Termux is your only terminal app — but it doesn’t carry over to other apps either.

KeyJawn is a system keyboard. It replaces Gboard, not your SSH client. You install it, enable it in Android settings, and the terminal keys work in every app. But the real difference is what it was built for: talking to AI agents via the CLI, where most of your input is natural language and the tools you use have slash commands, not vim keybindings.

The actual problem

Standard mobile keyboards are designed for texting. That’s fine until you need to send Ctrl+C to kill a stuck process, or Escape to exit a mode, or Tab to trigger autocomplete.

On Gboard:

• There’s no Escape key.
• There’s no Tab key.
• There’s no Ctrl key. No Ctrl+C to interrupt, no Ctrl+Z to suspend, no Ctrl+L to clear.
• Arrow keys are either hidden or nonexistent.
• Autocorrect turns git into got and npm into nap.

These aren’t edge cases anymore. They’re the first five minutes of trying to use Claude Code from your phone.

What KeyJawn does differently

The keyboard has a dedicated terminal row above the QWERTY layout: Esc, Tab, Ctrl, arrow keys, clipboard, mic, and upload. Always visible. No long-pressing, no layer hunting.

Voice input that streams. Most of what you type into an AI agent is natural language. “Fix the auth bug in login.ts.” “Refactor the dashboard component.” Thumb-typing that on a phone is slow. KeyJawn streams speech-to-text in real time — you see words appear as you say them, and you can edit the result before sending. For the 90% of CLI interaction that’s just talking to an agent in plain English, this is faster than typing.

SCP image upload. I SSH into Raspberry Pis from my phone constantly. When I need to share a screenshot with Claude Code — an error message, a UI bug, a terminal output — the normal process is: save the image, open a file manager, find an SCP client, configure the connection, transfer the file, switch back to the terminal, type the path. KeyJawn does it in one tap. Pick the photo, it SCPs to your server, and the remote file path gets typed at your cursor.

A system keyboard that works everywhere. Termius has terminal keys, but they’re inside the Termius app. Browser extensions add key rows, but they float above the page and break when you switch apps or the browser redraws. Hacker’s Keyboard is a system keyboard, but it’s a full desktop layout crammed onto a phone screen. KeyJawn installs as an Android input method — the terminal keys are part of the keyboard itself. They work in Termux, in Cockpit through Chrome, in JuiceSSH, in any app. You set it once and forget about it.

Customizable for how you actually work. I built this for my own workflow: SSHing into a Pi from a couch to run Claude Code sessions. The default key layout reflects that — the keys I reach for most are the ones I put in the terminal row. Slash command sets are configurable per tool (Claude Code, OpenClaw, Gemini CLI). Autocorrect toggles per app so you can leave it off for terminals and on for Slack. SSH hosts are saved so you can switch between machines without retyping credentials. The point is to fit the keyboard to your workflow instead of working around the keyboard.

Ctrl as a three-state toggle. Tap it once to arm it for one keypress (Ctrl+C). Long-press to lock it for a series of combos. Tap again to turn it off. Simple, and more useful than not having Ctrl at all.

Slash command picker. Claude Code, OpenClaw, and Gemini CLI all use slash commands (/help, /clear, /compact, /status). KeyJawn gives you a quick-pick popup instead of making you type them from memory. You can also create custom command sets for your own tools and workflows — add whatever commands you need, group them by project, toggle sets on and off.

Keyboard themes. Four color presets: the default dark, a light theme for outdoor use, true-black OLED for battery savings on AMOLED screens, and a green-on-black terminal theme for the aesthetic. The whole keyboard changes — keys, backgrounds, hints, accent colors.

Autocorrect is off by default. Autocorrect breaks web-based terminals because it uses setComposingText, which doesn’t map to the key events a shell expects. You can turn it on per app — keep it off for Termux, on for Slack.

Two versions

KeyJawn Lite is free. Full QWERTY keyboard with the terminal key row, number row, alt character popups, voice input, clipboard history (30 items), built-in slash commands, swipe gestures, per-app autocorrect, configurable quick key, color-coded extra row, haptic feedback, and shift/caps lock. Requires microphone permission for voice input. No network access, no tracking.

KeyJawn Full is $4, lifetime. Everything in the free version, plus SCP image upload, multi-host SSH management, encrypted credential storage, keyboard color themes (4 presets), an inline menu panel, clipboard pinning, custom slash command sets, and a tooltip toggle. No subscriptions, no ads, no tracking. One payment, done.

The code is MIT-licensed and on GitHub. Both versions are built from the same codebase — the full version unlocks features that need additional permissions (network, storage).

Try it

Download the APK and see if it fits your workflow. If something doesn’t work, file an issue. PRs are open.

Your phone keyboard shouldn’t be the reason you can’t use a terminal.