I built EZClaw (https://www.ezclaw.app) to solve a problem I kept running into: non-technical friends and family wanted their own AI assistant, but the setup barrier was too high. Self-hosting means Docker, API keys, cloud accounts, SSH — none of that is reasonable to ask of normal people.

EZClaw is managed hosting for OpenClaw (https://github.com/openclaw), an open-source personal AI assistant that lives in Telegram. You sign up, connect your Telegram, pick an AI model through OpenRouter (or use your openai/anthropic api key), and you're done. No cloud accounts, no API keys to manage, no servers to maintain.

What you get: - A personal AI assistant in Telegram that can search the web, manage files, run code, control smart home devices, and more - Choice of AI model (Claude, GPT-4, Gemini, Llama, etc.) via OpenRouter — you can bring your own OpenRouter/OpenAI/Anthropic key so you control model costs separately -- or rely solely on managed solution - Each user gets their own isolated container - The assistant remembers context across conversations

How it works technically: each user gets a dedicated OpenClaw instance running in a container. OpenClaw itself is an agent runtime — it connects to Telegram as your bot, routes to your chosen LLM via OpenRouter, and has a plugin system for tools (web search, browser automation, home assistant, etc.). EZClaw just handles the infra: provisioning, updates, monitoring, backups.

Pricing: $0.05/hr for the hosting. AI model costs are separate and go either through our openrouter infra or your OpenRouter/Anthropic/OpenAI account, so you have full visibility and control over that spend.

The underlying OpenClaw project is open source, so you can always self-host if you prefer. EZClaw is for people who'd rather not deal with that.

I'd love feedback on the product, pricing, or the approach in general. Happy to answer questions about the architecture.