OpenClaw

What is OpenClaw?

OpenClaw is a personal AI assistant you run on your own devices instead of a hosted service. It answers you on the chat apps you already use, keeps long-term memory of your preferences, and can act for you — controlling a browser, running commands, and firing automations. Its standout trait is self-improvement: it can write its own code to add new skills for tasks you ask for.

What is OpenClaw best for?

People who want a private, hackable assistant they own end to end — one that lives on their own Mac, Windows, or Linux machine, reaches them on the channels they already use, and runs real tasks rather than just chatting. It suits tinkerers and builders comfortable bringing their own LLM API key and doing light setup, not someone wanting a zero-config hosted app. If you want a similar server-side, always-on agent, Hermes Agent is a comparable self-hosted option.

What can OpenClaw do?

• Reach you on 20+ chat channels from one gateway — WhatsApp, Telegram, Slack, Discord, Signal, iMessage, Microsoft Teams, Matrix, and more
• Speak and listen, with wake words and a talk mode on macOS, iOS, and Android
• Render a Live Canvas — an agent-driven visual workspace you can watch and steer
• Act on your machine: control a browser, run shell commands, read and write files, and trigger webhooks and cron jobs
• Remember across sessions, learning your preferences and running proactive automations over time
• Write its own skills, autonomously generating code to add new capabilities on request
• Work with any model — connect Anthropic, OpenAI, or a local model and switch providers freely
• Install in one command and run as a background daemon, with a macOS menu-bar app plus iOS and Android companion nodes

Is OpenClaw free?

Yes — the software is free and MIT-licensed, so you can self-host it at no cost. But it ships no model of its own: you bring your own API key (Anthropic, OpenAI, or a local model) and pay that provider per use, or run a local model on your own hardware. Plan for ongoing LLM costs, plus an always-on machine to keep it running.

Where does OpenClaw fall short?

• It hands the assistant deep access to your system — file operations, shell commands, and browser control — so a misfire or prompt injection can do real damage. Independent reviewers flag the security model and advise caution; sandboxing it is on you.
• It isn’t free to run. You need an external LLM provider and ongoing per-call API spend (or a local model and the hardware to host it), plus a machine left on around the clock — some users buy a dedicated Mac Mini for it.
• It’s young, fast-moving software. First released in late 2025, it changes quickly, so expect rough edges, breaking changes between versions, and uncertain long-term support.

FAQ

Is OpenClaw open source? Yes — it’s released under the MIT License, so you’re free to self-host, modify, and redistribute it.

Do I need my own LLM API key? Yes. OpenClaw brings no model of its own. You connect a provider like Anthropic or OpenAI — or run a local model — and pay for that usage.

What do I need to run OpenClaw? A Mac, Windows, or Linux machine you can leave running, Node.js (v22.19+ or 24 recommended; the installer bundles it), and an LLM API key. A single npm command installs it and sets up a background daemon.

How is it different from ChatGPT? It runs on your own hardware, reaches you through your existing chat apps, and can actually act — controlling a browser, running commands, and writing its own skills — instead of only replying inside one web app.