Skip to content
Oday Bakkour
Back to Knowledge Hub

Clawdbot: The AI That Actually Does Things

Oday Bakkour profile photo
Oday Bakkour
4 min read
Share
Clawdbot: The AI That Actually Does Things

Clawdbot
The AI that actually does things.

Clears your inbox, sends emails, manages your calendar, checks you in for flights.
All from WhatsApp, Telegram, or any chat app you already use.

Quick Start

One-liner

bash.txt
curl -fsSL https://openclaw.ai/install.sh | bash

Works on macOS, Windows & Linux. The one-liner installs Node.js and everything else for you.

What It Does

Runs on Your Machine
Mac, Windows, or Linux. Anthropic, OpenAI, or local models. Private by default—your data stays yours.

Any Chat App
Talk to it on WhatsApp, Telegram, Discord, Slack, Signal, or iMessage. Works in DMs and group chats.

Persistent Memory
Remembers you and becomes uniquely yours. Your preferences, your context, your AI.

Browser Control
It can browse the web, fill forms, and extract data from any site.

Full System Access
Read and write files, run shell commands, execute scripts. Full access or sandboxed—your choice.

Skills & Plugins
Extend with community skills or build your own. It can even write its own.

Works With Everything

  • WhatsApp
  • Telegram
  • Discord
  • Slack
  • Signal
  • iMessage
  • Claude
  • GPT
  • Spotify
  • Hue
  • Obsidian
  • Twitter
  • Browser
  • Gmail
  • GitHub

View all 50+ integrations →

See what people built →

MacStories
"Clawdbot Showed Me What the Future of Personal AI Assistants Looks Like"
Federico Viticci

StarryHope
"The Lobster Takeover: Why Developers Are Buying Mac Minis to Run Their Own AI Agents"
Jim Mendenhall

https://docs.openclaw.ai/start/getting-started

https://openclaw.ai/

Update: Clawdbot is now OpenClaw

I wrote this post when the project was still called Clawdbot. It has been renamed twice since then, so here is what happened, based on the project's own lore page and the official announcement.

It was called Clawd from November 25, 2025. On January 27, 2026, Anthropic asked for a name change over trademark concerns, since Clawd sounds a lot like Claude. The community renamed it Moltbot that same day, after lobsters molting to grow. Moltbot never quite stuck, and within days the project became OpenClaw. The announcement post is dated January 29, 2026.

The code and the idea did not change. The old domains clawd.bot and molt.bot now redirect to openclaw.ai, and the repository is openclaw/openclaw on GitHub.

What else changed

  • Licence and funding: OpenClaw is MIT licensed and stewarded by the OpenClaw Foundation, a US 501(c)(3) non-profit. There is no paid tier, hosted service or token.
  • Install: the one-liner now comes from the new domain. Windows has its own PowerShell installer.
  • Runtime: it needs Node.js 24 or 26, and Node 26 is recommended. The installer sets Node up for you if needed.
  • Channels: the official site now lists 29 chat channels, including Microsoft Teams and Google Chat alongside WhatsApp, Telegram, Discord, Slack, Signal and iMessage.
  • Teams: one shared Gateway can now serve a team, with sessions others can open and steer.

Installing OpenClaw today

These are the current commands from the getting started guide. The last two confirm the Gateway is running (it listens on port 18789) and open the dashboard.

bash.txt
# macOS / Linux / WSL2
curl -fsSL https://openclaw.ai/install.sh | bash

# Windows (PowerShell)
# iwr -useb https://openclaw.ai/install.ps1 | iex

openclaw onboard --install-daemon
openclaw gateway status
openclaw dashboard

Security precautions before you give it access

The features I listed above, like full system access and browser control, are also its biggest risk. The OpenClaw README is direct about it: tools run on the host for the main session unless you configure sandboxing, and inbound messages should be treated as untrusted input. An email or web page the agent reads can contain instructions meant to hijack it. That is prompt injection, and no model fully prevents it.

The defaults are sensible. A normal install binds the Gateway to loopback, and DM channels reply to unknown senders with a pairing code instead of obeying them. You still need to do your part:

  • Run it on a separate machine, VM or user account, not the laptop with all your logins.
  • Give it dedicated email, calendar and API credentials with the smallest scope that works.
  • Turn on sandboxing and limit which tools each agent can use.
  • Run openclaw security audit after setup and after each config change.
  • Approve DM pairings yourself and keep group chats allowlisted.
  • Do not expose the Gateway to the internet without reading the official exposure runbook.

Read the full security analysis

I went much deeper in OpenClaw.ai: The Viral AI Agent Revolution and Its Critical Security Implications. It covers the prompt injection and data exposure research, why I do not recommend OpenClaw for non-experts, and a step-by-step hardened setup with a dedicated server, firewall rules, Cloudflare in front, credential isolation and Docker.

My view has not changed: OpenClaw shows where personal AI assistants are going, but treat it like a new employee with root access. Give it only what it needs, and check what it did.

Comments

Share your thoughts and join the conversation

Leave a Comment

Loading comments...
RELATED