IonClaw - The first AI Agent that runs everywhere - iOS, Android, tvOS, watchOS, Desktop, Servers

[paulocoutinhox]

Hi everyone 👋🚀

I want to present IonClaw 🐾⚡
The first AI Agent designed to run everywhere — powered by C++, llama.cpp, and local GGUF models.

🌐 Website: https://ionclaw.com/
⭐ GitHub: https://github.com/ionclaw/ionclaw

---

🐾 What is IonClaw?

IonClaw is an open-source C++ AI Agent Orchestrator built to bring AI agents to every device, from powerful servers to personal mobile devices and embedded hardware.

It runs natively on:

📱 iOS
🤖 Android
📺 tvOS
⌚ watchOS
💻 macOS
🪟 Windows
🐧 Linux
☁️ Servers / Cloud / VPS
🍓 Raspberry Pi

The goal is simple:

🤖 Put AI Agents everywhere.
On phones. On desktops. On servers. On TVs. On watches. On edge devices. On embedded systems.

---

🧠 Built with llama.cpp support

IonClaw supports local AI models using llama.cpp 🦙⚡

That means you can run GGUF models fully offline, directly on the device, without needing cloud APIs, external servers, or sending private data anywhere.

✅ Local inference
✅ Offline execution
✅ GGUF model support
✅ llama.cpp integration
✅ Private by design
✅ No API key required for local models
✅ Great for edge, mobile, desktop, and embedded AI

IonClaw also supports tools and agent workflows, so models can do more than just chat — they can execute tasks, use skills, interact with systems, and work as real AI agents.

---

🛠️ Tools, Skills and Marketplace

IonClaw was designed around extensibility.

It includes a Skills system 🧩 where agents can be extended with custom capabilities.

Skills can be used for things like:

🛠️ automation
🌐 web tasks
📂 file processing
🧠 reasoning workflows
🔌 device integrations
🤖 agent-to-agent workflows
📡 external tool connections
⚙️ custom native actions

There is also a Skills Marketplace 🛒✨ where people can discover, share, install, and improve skills for different use cases.

The idea is to make AI agents modular, reusable, and useful across many environments.

---

🤖 Multi-Agent Orchestration

IonClaw is not just a single chatbot.

It is an AI Agent Orchestrator.

That means you can run multiple agents with:

🤖 independent models
🛠️ independent tools
📂 separated workspaces
🧠 different roles
🔁 task routing
📡 agent-to-agent communication
🔒 sandboxed execution

This makes it possible to build real workflows where different agents collaborate together.

For example:

👨‍💻 one agent writes code
🧪 another agent tests it
📚 another agent reads documentation
🌐 another agent automates web actions
📦 another agent manages deployment tasks

---

🌐 Built-in Web Dashboard

IonClaw also includes a web-based control panel 📊

From the dashboard, you can:

👀 monitor agents in real time
📋 track tasks
🧩 edit and manage skills
📜 inspect logs
⚙️ control executions
🧠 manage agent behavior
📂 work with agent workspaces

This makes it easier to understand what your agents are doing instead of running everything as a black box.

---

🔌 MCP Support

IonClaw supports the Model Context Protocol as both:

🔗 MCP Client
🧩 MCP Server

This allows IonClaw to connect with MCP-compatible tools, platforms, and workflows.

The goal is interoperability: agents should be able to communicate with tools, systems, and other agents in a structured way.

---

🔐 Offline, Private and Native

One of the biggest goals of IonClaw is privacy.

With llama.cpp and local GGUF models, your AI agent can run directly on your device.

No cloud required.
No external API required.
No data leaving your device.
No Python runtime required.
No Node.js runtime required.
No heavy dependency chain.

Just native C++ performance ⚡

---

🚀 Why I built this

Most AI agent platforms are focused on cloud servers, Python environments, or browser-based systems.

But I believe AI agents should run everywhere:

📱 in your phone
⌚ on your watch
📺 on your TV
💻 on your computer
☁️ on your server
🍓 on your Raspberry Pi
🔌 inside embedded devices
🏠 in home automation
🤖 in robotics
🏭 in industrial systems
🌎 at the edge

AI agents should not be limited to one environment.

---

🤝 Looking for testers and contributors

IonClaw is free and open source ❤️

I would love help from the community to test it on different devices, platforms, and hardware setups.

If you have:

📱 Android devices
🍎 iPhones / iPads
⌚ Apple Watch
📺 Apple TV
💻 Linux / macOS / Windows machines
🍓 Raspberry Pi boards
☁️ servers or VPS environments
🦙 experience with llama.cpp / GGUF models

Your feedback would be extremely valuable.

You can help by:

⭐ starring the repository
🐛 reporting bugs
🧪 testing on your devices
💡 suggesting features
🧩 creating skills
📚 improving documentation
🔧 contributing code
🦙 testing different llama.cpp models
🚀 helping bring AI agents to more platforms

---

🌍 The vision

AI agents should not live only in the cloud.

They should run locally.
They should run privately.
They should run on real devices.
They should work offline.
They should be extensible.
They should be open source.
They should be everywhere.

Let’s put AI Agents everywhere. 🤖🌍🔥

Thanks for reading!
Feedback, testing, ideas, and contributions are very welcome 🙏🚀

[paulocoutinhox]

Running on Apple TV: https://www.youtube.com/watch?v=YwnjUi5Mj2o

[CISC]

Does it run in Doom?

[paulocoutinhox]

SURE SEE