OpenClaw on Android - Setup Guide¶
Repo: AidanPark/openclaw-android
Stars: 1,649 | Author: Aidan Park
Requirements: Android 7.0+, Termux
Skills.sh: npx skills add reason-machines/hermes-skills --skill openclaw-android-setup
What It Is¶
Run OpenClaw - the autonomous AI agent framework - directly on Android without a full Linux distribution. The standard approach requires proot-distro + Debian/Ubuntu (700MB-1GB overhead). This project installs just the glibc dynamic linker (ld.so), letting Node.js and OpenClaw run natively on Android's Linux kernel through Termux.
Architecture Comparison¶
Standard approach (heavy):
Linux Kernel → Android/Bionic libc → Termux → proot-distro → Debian/Ubuntu → glibc → Node.js → OpenClaw
This project (lightweight):
Linux Kernel → Android/Bionic libc → Termux → glibc → Node.js → OpenClaw
Prerequisites¶
- Android 7.0+ (API level 24+)
- Termux installed (from F-Droid, NOT Google Play - the Play Store version is outdated)
- ~500MB free storage (vs 1.5GB+ for the proot-distro approach)
- Internet connection for initial setup
Installation¶
1. Install Termux¶
Download from F-Droid (recommended):
https://f-droid.org/packages/com.termux/
⚠️ Do NOT use the Google Play Store version - it's outdated and missing critical APIs.
2. Run the One-Command Setup¶
Open Termux and run:
curl -fsSL https://raw.githubusercontent.com/AidanPark/openclaw-android/main/setup.sh | bash
This single command: - Installs required Termux packages (git, curl, nodejs, etc.) - Downloads and sets up the glibc dynamic linker - Installs Node.js with glibc compatibility - Clones and configures OpenClaw - Creates startup scripts
3. Start OpenClaw¶
# Start OpenClaw (first run will prompt for configuration)
openclaw start
# Or use the convenience script
~/openclaw-android/start.sh
Configuration¶
First-Run Setup¶
On first launch, OpenClaw will prompt for:
1. API provider - OpenAI, Anthropic, DeepSeek, local model, etc.
2. API key - Your provider's API key
3. Model - Default model selection (e.g., claude-sonnet-4-20250514)
4. Workspace directory - Where OpenClaw stores sessions and configs
Config File Location¶
~/.openclaw/config.yaml
Edit directly:
nano ~/.openclaw/config.yaml
Example Configuration¶
provider: anthropic
model: claude-sonnet-4-20250514
api_key: ${ANTHROPIC_API_KEY}
workspace: ~/openclaw-workspace
skills_dir: ~/openclaw-skills
Usage¶
Basic Commands¶
# Start OpenClaw in interactive mode
openclaw
# Run a one-shot task
openclaw run "Summarize the latest AI research papers"
# Start in daemon mode (runs in background)
openclaw daemon start
# Check daemon status
openclaw daemon status
# Stop daemon
openclaw daemon stop
# List installed skills
openclaw skills list
# Install a skill from skills.sh
openclaw skills install reason-machines/hermes-skills --skill hermes-agent-framework
Running on Old Android Phones¶
This setup is particularly well-suited for repurposing old Android phones: - Samsung Galaxy S8+ (2017) - runs OpenClaw comfortably - Google Pixel 2 (2017) - known working configuration - Any phone with 4GB+ RAM and Android 7.0+
Battery Optimization¶
For long-running agents:
# Disable battery optimization for Termux (Android settings)
# Settings → Apps → Termux → Battery → Unrestricted
# Or via ADB:
adb shell dumpsys deviceidle whitelist +com.termux
Skills Available¶
Once OpenClaw is running, install Hermes-compatible skills:
# Agent infrastructure
openclaw skills install reason-machines/hermes-skills --skill hermes-agent-framework
openclaw skills install reason-machines/hermes-skills --skill hermes-agent-self-evolution
# Communication
openclaw skills install reason-machines/hermes-skills --skill hermesclaw-wechat-multi-agent
# Security
openclaw skills install prompt-security/clawsec --skill soul-guardian
Troubleshooting¶
"glibc not found" error¶
# Re-run the glibc setup
cd ~/openclaw-android
./scripts/setup-glibc.sh
"Termux packages outdated"¶
# Update Termux packages
pkg update && pkg upgrade
Node.js version mismatch¶
# The setup installs a specific Node.js version for glibc compatibility
# To reinstall:
cd ~/openclaw-android
./scripts/install-node.sh
Storage space issues¶
# Check free space
df -h /data/data/com.termux/files/home
# Clean up unused packages
pkg clean
Wake lock / sleep issues¶
Android may kill Termux background processes to save battery:
# Acquire a wake lock in Termux
termux-wake-lock acquire
# Release when done
termux-wake-lock release
Security Notes¶
- API keys are stored in
~/.openclaw/config.yaml- set appropriate file permissions - Consider using environment variables:
provider: anthropicwithapi_key: ${ANTHROPIC_API_KEY} - The Android device should have screen lock enabled
- For production use, consider a dedicated device with no personal data
CorpusIQ Use Cases¶
| Use Case | How |
|---|---|
| Mobile monitoring agent | Run an OpenClaw agent on an old phone to monitor social mentions 24/7 |
| Edge AI deployment | Repurpose old Android devices as low-power Hermes-compatible agent nodes |
| Field testing | Test agent skills on mobile before deploying to production servers |
| Personal assistant | Always-on agent on your secondary phone for research, scheduling, reminders |
| Learning platform | Low-cost Hermes/OpenClaw sandbox for experimenting with agent configurations |
Comparison: Android vs Other Deployment Options¶
| Platform | Setup Time | Storage | RAM Required | Power Draw |
|---|---|---|---|---|
| Android (this project) | 5 min | 500MB | 2GB | 3-5W |
| Linux server (cloud) | 2 min | 2GB | 4GB | N/A |
| Raspberry Pi 5 | 15 min | 8GB | 4GB | 5-8W |
| Mac Mini | 5 min | 5GB | 8GB | 15-20W |
Setup guide by CorpusIQ. Project by Aidan Park. OpenClaw framework by the open-source community.
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.