Getting Started with OpenHoof

OpenHoof is an uncensored, self-hosted AI agent platform. It runs as a local daemon on your machine and connects to any chat app, giving you a personal AI with persistent memory, web browsing, code execution, and zero restrictions.

What is OpenHoof?

OpenHoof is an AI gateway that sits between you and any large language model. Instead of chatting through a web UI controlled by someone else, you run OpenHoof locally and interact with your AI through the apps you already use — Telegram, Discord, WhatsApp, Slack, Signal, iMessage, or LINE.

Your AI agent has full access to your machine. It can browse the web, execute shell commands, read and write files, schedule tasks, and remember everything you've ever discussed. There are no guardrails, no content filters, and no usage limits beyond what your model provider charges.

Quick Start

Get running in under two minutes:

# Install globally
npm i -g openhoof

# Initialize your config
openhoof init

# Start the daemon
openhoof start

After starting, OpenHoof runs as a background daemon. You'll get a local web UI at http://localhost:3777 where you can chat immediately. To connect a chat app, see the Channels guide.

Core Concepts

The Daemon

OpenHoof runs as a persistent background process. It listens for messages from connected channels, routes them to your configured AI model, and streams responses back. The daemon also manages memory, scheduled tasks, and skill execution.

# Check daemon status
openhoof status

# View logs in real-time
openhoof logs -f

# Stop the daemon
openhoof stop

Channels

Channels are the interfaces through which you talk to your AI. Each channel connects to a different messaging platform. You can have multiple channels active simultaneously — message your AI on Telegram while it also monitors a GitHub. See Channels for setup instructions.

Skills

Skills extend your agent's capabilities. They're plugins that give your AI new tools — web browsing, code execution, image generation, file management, and more. Install community skills or write your own. See Skills for details.

Memory

OpenHoof maintains persistent memory using vector embeddings stored locally. Your AI remembers past conversations, user preferences, and facts across sessions. Memory is searchable and can be manually managed. See Memory for the full guide.

Agents

An agent is the AI model configuration that powers your responses. OpenHoof supports Claude, GPT, Gemini, Grok, and local models via Ollama. You can configure multiple agents for different channels or use cases. See Agents.

System Requirements

• Node.js v18 or later
• OS: macOS, Linux, or Windows (WSL recommended)
• RAM: 512MB minimum (more if running local models via Ollama)
• Disk: ~200MB for OpenHoof + memory storage

Project Philosophy

OpenHoof exists because AI should work for you, not the other way around. Commercial AI products decide what you can and can't ask. They log your conversations. They refuse to help with legitimate tasks because of overly broad content policies.

OpenHoof is different. It's your machine, your data, your rules. The platform is model-agnostic — swap providers any time. Your conversation history and memory stay on your disk, never uploaded anywhere. And there are no artificial restrictions on what your AI can discuss or do.