Zecel Manatad

Posted on May 14

Running Claude Code, Ollama, and OpenClaw on Android using Termux + Ubuntu (2026 Guide)

#ai #claude #openclaw

This guide documents a real setup process of turning an Android phone into a portable AI development environment using Termux, Ubuntu (proot), Node.js, Ollama, and OpenClaw.

The goal: run modern AI coding tools on a mobile device without root.

🧱 1. Base Setup: Termux + Ubuntu

We start by installing Termux and setting up a Linux environment.

Install Termux packages:

pkg update && pkg upgrade -y
pkg install proot-distro git curl wget -y

Install Ubuntu:

proot-distro install ubuntu
proot-distro login ubuntu

Now we have a full Linux environment running on Android.

⚙️ 2. Installing Node.js (Critical Step)

Many modern AI tools require Node.js 22+.

Initial issue encountered:

«Node.js version mismatch (required 22.12+, installed 20.x)»

Fix:

Install Node.js using NVM (recommended for Android):

curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.7/install.sh | bash
source ~/.bashrc

nvm install 22
nvm use 22
nvm alias default 22

Verify:

node -v
npm -v

🤖 3. Installing Ollama (Local AI Models)

Ollama allows running local LLMs directly on the device.

Install:

curl -fsSL https://ollama.com/install.sh | sh

Start server:

ollama serve

Run a model:

ollama run qwen2.5-coder:3b

For mobile devices, lightweight models are recommended:

qwen2.5-coder:3b
phi4-mini
gemma3:4b

🧠 4. Installing Claude Code

Claude Code is an AI coding CLI tool.

Install:

npm install -g @anthropic-ai/claude-code

Issue encountered:

“native binary not installed”
caused by Android/proot incompatibility

Fix:

Run inside Ubuntu environment only, not raw Termux.

⚠️ 5. Fixing npm Installation Errors

A major issue appeared:

ENOENT rename /root/.npm/_cacache/tmp
Invalid response body from registry

Fix:

rm -rf ~/.npm
npm cache clean --force
npm config set cache /tmp/npm-cache
mkdir -p /tmp/npm-cache

Then retry installation.

🔧 6. Installing OpenClaw (AI Agent)

OpenClaw is an AI agent system that can automate coding tasks.

Install:

npm install -g openclaw

Issue encountered:

Node version requirement mismatch (Node 22.12 required)

Fix:

Upgrade Node.js to version 22 using NVM.

⚙️ 7. Configuring OpenClaw

Create config file:

nano ~/.openclaw/config.json

Example configuration using Ollama:

{
"agent": {
"model": "ollama/qwen2.5-coder:3b"
},
"providers": {
"ollama": {
"base_url": "http://127.0.0.1:11434"
}
},
"features": {
"allow_shell": true,
"allow_file_edit": true
}
}

🚀 8. Running the Full Stack

Start Ollama:

ollama serve

Start OpenClaw:

openclaw chat

Now the system can:

analyze code
run shell commands
use local AI models
automate development tasks

🧩 9. Key Issues Encountered

Node.js version mismatch

Fixed using NVM (Node 22)

npm cache corruption

Fixed by clearing ~/.npm and using /tmp cache

Android proot filesystem issues

Fixed by avoiding Termux-native installs and using Ubuntu environment

Ollama connectivity issues

Fixed using 127.0.0.1 instead of localhost

🧠 Final Architecture

Android (Termux)
↓
Ubuntu (proot)
↓
Node.js 22 + NVM
↓
Ollama (local AI)
↓
OpenClaw (AI agent)
↓
Claude API (optional cloud intelligence)

🔥 Conclusion

Android can be transformed into a full AI development workstation using Termux and Ubuntu without root.

While some tools (like OpenClaw and Claude Code) require workarounds due to Linux/Android differences, a hybrid setup of local + cloud AI makes the system powerful and practical.

This setup is ideal for: