malikclaw

mcp
Guvenlik Denetimi
Gecti
Health Gecti
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 14 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested
Purpose
This is an ultra-lightweight AI agent and assistant designed to run on low-cost hardware (like $10 single-board computers) and Android devices. It facilitates an agentic loop (Plan, Act, Observe, Reflect) while emphasizing low memory usage and offline privacy.

Security Assessment
The overall risk is Low. The light code scan (12 files) found no dangerous patterns, hardcoded secrets, or excessive permission requests. The tool naturally requires network access to communicate with external AI models via user-provided API keys, but the configuration is stored locally. The primary security consideration is the installation method: it relies on piping remote scripts directly into the shell (`curl | bash` and `irm | iex`). While standard for many open-source projects, this method bypasses manual review and requires trusting the repository maintainer. It is highly recommended to download and inspect the installation scripts before executing them.

Quality Assessment
Project quality is Good. It is actively maintained, with the most recent push occurring today. It is fully licensed under the permissive MIT standard. The project currently has 14 GitHub stars, indicating a small but growing level of community trust and early adoption. The repository includes multiple multilingual READMEs and a comprehensive quick-start guide, reflecting a strong focus on accessibility and documentation.

Verdict
Safe to use, but always inspect the `install.sh` and `install.ps1` scripts before running them on your system.
SUMMARY

Ultra-lightweight AI Agent & Assistant for $10 SBCs & Android. Urdu-first, Go-based, and privacy-focused. 99% smaller memory than typical gateways.

README.md
MalikClaw AI Agent running on $10 Linux SBC

MalikClaw 🦅

Ultra-Efficient Personal AI Assistant

Production-Grade • <10MB RAM • <1s Boot • $10 Hardware • Built for the World

Build Status Golang 1.21+ Platform Support MIT License
GitHub stars GitHub forks GitHub issues Official Website

Quick StartFeaturesInstallationDemoDocumentationCommunity

اردو | 日本語 | Português | Tiếng Việt | Français | Deutsch | Español | العربية | English


🚀 Quick Start

One-Command Installation

Linux/macOS:

curl -fsSL https://raw.githubusercontent.com/AbdullahMalik17/malikclaw/main/install.sh | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/AbdullahMalik17/malikclaw/main/install.ps1 | iex

Docker:

docker run -d --name malikclaw -p 18790:18790 -v ~/.malikclaw:/root/.malikclaw ghcr.io/abdullahmalik17/malikclaw:latest

Setup (5 Minutes)

  1. Run onboarding: malikclaw onboard
  2. Add API keys: Edit ~/.malikclaw/config.json
  3. Test: malikclaw agent -m "Hello!"

📖 Full Guide: PRODUCT.md | QUICK_REFERENCE.md


✨ Features

🤖 Production-Grade Agent Loop

NEW: Advanced agentic architecture with full execution cycle:

PLAN → ACT → OBSERVE → REFLECT → MEMORY UPDATE
  • Intelligent Planning: Goal decomposition into actionable steps
  • Resilient Execution: Retry logic with exponential backoff and circuit breaker
  • Smart Observation: Result capture, normalization, and confidence scoring
  • Deep Reflection: Success/failure evaluation with lessons learned
  • Persistent Memory: Episode storage with search and analytics

📖 Technical Docs: pkg/agent/agentloop/README.md

🌍 Multi-Language Support

  • Localized Interface: Urdu, English, Japanese, French, Portuguese, Vietnamese, and growing
  • RTL Support: Native Right-to-Left support for Urdu/Arabic users
  • Global Community: Contributors from 6+ continents

🪶 Ultra-Lightweight

  • <10MB RAM — 99% smaller than typical AI agents
  • <1s Boot — 400X faster startup than Python-based alternatives
  • $10 Hardware — Runs on Orange Pi Zero, Raspberry Pi Zero, old Android phones
  • Cross-Platform — Linux, macOS, Windows, Docker

📱 Mobile Operation

  • ADB Control: Screenshot, tap, swipe, type on Android devices
  • Termux Support: Run directly on Android without root
  • Second Life: Give your old phone a purpose as an AI assistant

💼 Business Integration

  • Gmail: Send and manage emails
  • Calendar: Schedule and manage events
  • MCP Support: Odoo, custom business integrations
  • Multi-Channel: Telegram, Discord, WhatsApp, Matrix, QQ, DingTalk, LINE, WeCom

🛠️ Self-Evolution

  • Guardian Engine: Autonomous code improvement
  • Skill System: Extensible via plugins
  • Web Search: DuckDuckGo, Tavily, Brave, Perplexity, SearXNG

📊 Performance Comparison

Metric OpenClaw NanoBot MalikClaw
Language TypeScript Python Go
RAM Usage >1GB >100MB <10MB
Boot Time (0.8GHz) >500s >30s <1s
Hardware Cost $599 ~$50 $10
Privacy Cloud Local 100% Local
Performance comparison chart

📦 Installation

System Requirements

Platform Minimum Recommended
Linux SBC Orange Pi Zero ($10) Raspberry Pi 4
Android 2GB RAM, Termux 4GB RAM
Desktop 2GB RAM, any OS 4GB RAM
Docker 512MB RAM 1GB RAM

Installation Methods

1. One-Command Install (Recommended)

See Quick Start above.

2. Build from Source

git clone https://github.com/AbdullahMalik17/malikclaw.git
cd malikclaw

# Build for current platform
make build

# Build and install
make install

# Build for multiple platforms
make build-all

3. Docker

# Minimal (Alpine-based)
docker compose -f docker/docker-compose.yml up

# Full-featured (Node.js 24 for MCP support)
docker compose -f docker/docker-compose.full.yml up

4. Package Managers

Homebrew (macOS):

brew install malikclaw

Scoop (Windows):

scoop install malikclaw

AUR (Arch Linux):

yay -S malikclaw

🎮 Demo

Interactive Demo Script

Run the comprehensive demo showcasing all capabilities:

curl -fsSL https://raw.githubusercontent.com/AbdullahMalik17/malikclaw/main/demo.sh | bash

Demo 1: Phone Control (ADB)

Prerequisites:

  • Android device with USB debugging enabled
  • ADB installed (sudo apt install android-tools-adb)

Commands:

# Take a screenshot
malikclaw agent -m "Take a screenshot of my phone"

# Tap at coordinates
malikclaw agent -m "Tap at position 500, 1000 on my phone"

# Type text
malikclaw agent -m "Type 'Hello World' on my phone"

# Open an app
malikclaw agent -m "Open WhatsApp on my phone"

# Swipe gesture
malikclaw agent -m "Swipe up on my phone screen"

Demo 2: Email (Gmail)

Prerequisites:

  • Gmail account
  • Gmail API enabled in config

Commands:

# Send an email
malikclaw agent -m "Send an email to [email protected] with subject 'Meeting' and body 'Let's meet at 3pm tomorrow'"

# Check recent emails
malikclaw agent -m "Show my 5 most recent emails"

# Search emails
malikclaw agent -m "Find emails from last week about 'project'"

# Reply to email
malikclaw agent -m "Reply to the last email with 'Thanks, I'll review it'"

Demo 3: Web Search

Prerequisites:

  • Internet connection
  • DuckDuckGo enabled by default (no API key needed)

Commands:

# Get current weather
malikclaw agent -m "What's the weather in Lahore today?"

# Search for news
malikclaw agent -m "What are the latest AI news today?"

# Research topic
malikclaw agent -m "Research quantum computing breakthroughs in 2025"

# Get documentation
malikclaw agent -m "Fetch the Go 1.25 release notes"

# Find tutorials
malikclaw agent -m "Find Go programming tutorials for beginners"

🖥️ Web Interface

Access the modern web UI at: http://localhost:18790

Features:

  • Real-time chat interface
  • Status dashboard (agent status, memory usage, tools, uptime)
  • Demo task shortcuts
  • Responsive design (mobile-friendly)
  • Dark theme

Start Web UI:

malikclaw gateway

Or with Docker:

docker run -d -p 18790:18790 -v ~/.malikclaw:/root/.malikclaw ghcr.io/abdullahmalik17/malikclaw:latest

⚙️ Configuration

Basic Config

Edit ~/.malikclaw/config.json:

{
  "model_list": [
    {
      "model_name": "gpt-4o-mini",
      "model": "openai/gpt-4o-mini",
      "api_key": "sk-your-api-key-here"
    }
  ],
  "tools": {
    "web": {
      "duckduckgo": { "enabled": true },
      "tavily": { "enabled": false }
    },
    "shell": { "enabled": true },
    "file": { "enabled": true }
  },
  "performance": {
    "low_memory_mode": true,
    "max_concurrent_tasks": 2
  }
}

Get API Keys

LLM Providers:

Web Search (Optional):


🎯 Use Cases

1. Personal Assistant

  • Manage emails and calendar
  • Set reminders and alarms
  • Search for information
  • Control smart home devices

2. Developer Tool

  • Code generation and review
  • File operations and search
  • Git operations
  • Documentation lookup
  • TODO comment tracking

3. Mobile Automation

  • Automate repetitive phone tasks
  • Take screenshots and extract text
  • Send messages via WhatsApp
  • Control apps remotely

4. Edge AI

  • Run on low-power hardware
  • Offline-capable with local models
  • Privacy-first (data stays local)
  • 24/7 operation with minimal power

📚 Documentation

Document Description Audience
PRODUCT.md Quick start guide (5-min setup) New users
QUICK_REFERENCE.md Command cheat sheet All users
README.product.md Product overview General
pkg/agent/agentloop/README.md Agent loop architecture Developers
CONTRIBUTING.md Contribution guidelines Contributors
ROADMAP.md Development roadmap Community

🔧 Troubleshooting

Common Issues

"Command not found" after installation:

export PATH="$HOME/.local/bin:$PATH"
echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.bashrc
source ~/.bashrc

"API key not configured":

  1. Open config: malikclaw config edit
  2. Add your API key to model_list
  3. Save and retry

"ADB device not found":

# Check USB debugging is enabled on phone
# List connected devices
adb devices

# If empty, try:
adb kill-server
adb start-server
adb devices

"Out of memory" on low-RAM devices:

{
  "performance": {
    "low_memory_mode": true,
    "max_concurrent_tasks": 1,
    "gc_interval_seconds": 180
  }
}

📖 More Help: PRODUCT.md#troubleshooting


🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

Ways to Contribute

  • 🐛 Report bugs
  • ✨ Add features
  • 📝 Improve documentation
  • 🌍 Translate to your language
  • 💡 Share use cases
  • 🔌 Build skills/plugins

Development Setup

git clone https://github.com/AbdullahMalik17/malikclaw.git
cd malikclaw

# Download dependencies
make deps

# Run tests
make test

# Build
make build

# Run linters
make lint

📢 Community

Join our global community of developers and users:


📄 License

MIT License — See LICENSE file for details.


🌍 Built for the World

MalikClaw is proudly used by developers and organizations worldwide:

  • 🌏 Asia: Pakistan, India, Japan, China, Vietnam, Philippines
  • 🌍 Europe: Germany, France, UK, Netherlands, Poland
  • 🌎 Americas: USA, Brazil, Canada, Mexico, Argentina
  • 🌍 Africa: Nigeria, Egypt, South Africa, Kenya
  • 🌏 Oceania: Australia, New Zealand

Contributors from 20+ countries and growing every day!


🦅 Built with ❤️ for the World

Empowering everyone with efficient AI — anywhere, anytime!


📢 Recent Updates

March 2026 - Production-Grade Agent Loop

NEW: Complete agent loop implementation with:

  • Planner: Goal decomposition with LLM and heuristic planning
  • Executor: Tool execution with retry, backoff, and circuit breaker
  • Observer: Result capture, normalization, and confidence scoring
  • Reflector: Success evaluation with lessons learned
  • Memory: Persistent episode storage with search capabilities

Files Added:

  • pkg/agent/agentloop/ - Main orchestrator and configuration
  • pkg/agent/planner/planner.go - Enhanced planning system
  • pkg/agent/executor/executor.go - Resilient execution engine
  • pkg/agent/observer/observer.go - Observation capture
  • pkg/agent/reflector/reflector.go - Reflection engine
  • pkg/agent/memory/memory.go - Memory management

Product Enhancements:

  • One-command installers for Linux/macOS and Windows
  • Interactive demo script showcasing phone, email, and search capabilities
  • Modern web UI with chat interface and status dashboard
  • Comprehensive documentation (PRODUCT.md, QUICK_REFERENCE.md)

🦅 Built with ❤️ for South Asian developers

آگے بڑھو، ملک کلاؤ! (Let's Go, MalikClaw!)

Yorumlar (0)

Sonuc bulunamadi