Run your own AI agent, step by step.

What a local-first AI agent is, what actually stays on your machine, and how to run one on Mac or Windows with CoWork OS, Ollama or MLX. Free and open source.

Use Ollama as the brain of a full AI agent, not just a chat window. Step-by-step setup in CoWork OS, model picks, context length and offline use.

Set up a local AI agent on an Apple Silicon Mac with MLX-LM: requirements, install commands, recommended Qwen models and how CoWork OS starts the MLX server for you.

Connect MCP servers to a free, open-source desktop app that works with any model, local or cloud. Add stdio or remote OAuth servers, browse the registry, or expose CoWork as an MCP host.

Run your own AI agent as a Telegram bot from your computer. Create a bot with BotFather, connect it to CoWork OS, lock it down with pairing codes and use any model.

Add a self-hosted AI agent to Slack that runs on your own machine or server. Create the Slack app, set scopes and Socket Mode tokens, and connect it to CoWork OS.

Talk to your own AI agent on WhatsApp by linking it like WhatsApp Web, no Business API account required. Setup steps, self-chat mode, and when to use the Business API instead.

Exactly what leaves your machine when you use CoWork OS: opt-in usage counts, update checks and the model calls you configure. What is stored locally and how it is encrypted.

An honest look at OpenClaw security: what its own docs warn about, the ClawHub malicious-skill problem, how to harden a gateway, and safer alternatives.