I Can't Write Code. But I Built a 100,000-Line Terminal IDE on My Phone.

#ai #opensource #android #vibecoding

I can't write code.

I'm not an engineer. I've never written a line of TypeScript. I have no formal training in computer science.

But I built a 100,000-line terminal IDE — by talking to AI.

Every architectural decision is mine. The code is not. It was created through conversation with Claude Code, running inside Termux on a Samsung Galaxy Z Fold6. No desktop. No laptop. Just a foldable phone and an AI that can execute commands.

Today I'm releasing it as open source.

GitHub: github.com/RYOITABASHI/Shelly

The Problem

You're running Claude Code in the terminal. It throws an error. You copy it. You switch to ChatGPT. You paste. You ask "what went wrong?" You copy the fix. You switch back. You paste. You run it.

Seven steps. Every single time.

The terminal and the chat live in different worlds. You are the copy-paste bridge between them.

The Solution

Shelly puts Chat and Terminal side by side — and connects them with AI.

Say "fix the error on the right". Shelly reads the terminal output, understands the error, and generates an executable command. Tap ▶ Run and the fix lands directly in the Terminal.

No copy. No paste. No tab switching.

What makes it different

Native terminal emulator. Not a WebView. Kotlin code derived from Termux's terminal-emulator library, rendering on an Android Canvas. Connected via a C helper (forkpty()) over TCP localhost. As far as I know, this is the only React Native app with an embedded native terminal emulator.

12 AI agents. @claude, @gemini, @codex, @cerebras, @local, @perplexity, @team, @plan, @arena, @actions. The input router automatically selects the best AI for the task, or you choose with @mentions.

It runs on your phone. Not in the cloud. Not on a desktop. Shelly + Termux gives you a full development environment in your pocket.

Key Features

Cross-pane intelligence — AI reads terminal output, suggests fixes, one-tap execution
ActionBlock — Code blocks in AI responses have [▶ Run] buttons
Approval Proxy — Terminal [Y/n] prompts become native chat buttons
Auto-savepoint — Game-like save system. Every change auto-committed. Revert with one tap.
Plan Mode — AI generates step cards. Execute or skip each step.
Arena Mode — Same prompt, two AIs, blind comparison. Vote for the better one.
Click-to-Edit — Tap any element in the preview panel, tell the AI what to change
VoiceChain — Speak → execute → TTS response. Hands-free development.
GitHub integration — AI suggests when to push. @actions sets up CI/CD without YAML.
5-level command safety — Every command is risk-assessed before execution
Local LLM support — Run models on-device via llama.cpp. Fully offline capable.

The keyboard too

The keyboard you see in the screenshots is Nacre — an 11,000-line Android IME I built (also through AI) to solve the input problem on mobile.

github.com/RYOITABASHI/Nacre

Built with

Layer	Technology
Framework	Expo 54 / React Native 0.81
Language	TypeScript + Kotlin + C
Terminal	Native emulator (Termux-derived) + Direct PTY (C, forkpty)
State	Zustand
i18n	900+ keys, English/Japanese

Why I built this

I'm a creative director. I wanted to use Claude Code on my phone, but Termux was too intimidating. So I made a chat interface that hides the complexity while keeping the full power.

Then I realized the real problem wasn't the terminal — it was the gap between the terminal and the AI. So I connected them.

Then the WebView kept dying every time I switched apps. So I directed the AI to replace the entire rendering layer with a native terminal emulator written in Kotlin and C.

100,000 lines later, I still can't write code. But I can describe what I need, and the AI builds it.