ZeroClaw

ZeroClaw is an open-source, self-hosted agent framework from ZeroClaw Labs that runs an AI assistant entirely on your own machine. Instead of shipping conversations off to a hosted SaaS, you install it with a single shell script, point it at whichever LLM backend you prefer (Claude, OpenAI or a local model via something like Ollama or llama.cpp), and wire it into the messaging platforms you already use — Telegram, Discord, WhatsApp and Slack are the headline integrations. The agent gets persistent memory across sessions, can be extended with tools, and multiple agents can be run side by side, so a single install can host a personal-assistant bot in Telegram, a moderator bot in Discord and a WhatsApp responder simultaneously. Because everything runs locally by default, messages, files and any long-term memory stay on the host machine unless you explicitly grant a cloud LLM access to a prompt. That makes it a natural fit for people who want the ergonomics of a chat-based agent (talk to your bot from your phone, get answers back on whatever platform you're already in) without handing their data to a third-party assistant vendor. Typical workflows include always-on personal assistants that summarise inbound messages, home-automation controllers wired to local scripts, private research assistants that can browse and take notes, and small-team ops bots living inside a Discord or Slack workspace. It is a developer-oriented product — install is CLI-driven, configuration is text-file based, and getting real value out of it assumes you can bring your own LLM credentials or run a local model — but the framework itself asks for no subscription and no cloud account.