Codista

Clawdbot AI: 5 minutes to install, 5 seconds to get hacked

January 25, 2026•5 min read
Featured image for article: Clawdbot AI: 5 minutes to install, 5 seconds to get hacked - The Jarvis that Siri never was. Integrates with WhatsApp, Telegram, Gmail and more. Use it without becoming a security s...
#ai#clawdbot#security#claude

Clawdbot was released last November but went viral in January 2026. Why the hype? Let me explain...

68,596stars
clawdbot/clawdbot star history chart

Chart by star-history.com • Click for interactive view

In this article:

  1. What is Clawdbot AI?
  2. How to install Clawdbot?
  3. How to improve Clawdbot security?
  4. Automatic integrations
  5. Clawdbot vs Claude Cowork
  6. Cheatsheet

What is Clawdbot AI?

It's an application that connects AI to your digital life. Works as a free AI assistant for WhatsApp and a Telegram bot.

It runs on your computer, connects to your favorite AI provider (Anthropic, OpenAI, Google, Microsoft, Ollama, etc.) and has claws that manage your digital tools: WhatsApp, Telegram, Discord, Slack, Microsoft Teams, Google Chat, iMessage, Gmail, Google Calendar, Google Chrome and, most importantly: your terminal.

Supports today's best models: Opus 4.5, GPT 5.2, Gemini 3, Haiku and Nano Banana Pro

It's Iron Man's Jarvis with access to your WhatsApp, emails, calendar, browser, notes and terminal.

A bridge between natural language and the digital world. Revolutionary because it's what Siri never managed to be.

How to install Clawdbot?

Prerequisite: Node.js version 22+

The official guide is available on the GitHub repository: clawdbot/clawdbot

Quick, easy and secure (offline)

Install it on your computer, don't open any ports, control it via web dashboard or Telegram. This eliminates 95% of the risk of being hacked.

npm install -g clawdbot@latest clawdbot onboard --install-daemon

Follow the CLI instructions.

My recommendations

Set up Telegram first and use it there, don't start with WhatsApp.

Telegram integration has some advantages:

  • Security through parity
  • Distinction between your messages and the bot's

When integrating with WhatsApp the standard way (on your number), the bot can only send messages as if it were you, causing a poor experience:

Clawdbot demo on WhatsApp

With Telegram, it's much better! Clawdbot demo on Telegram

How to improve Clawdbot security?

Under no circumstances ever expose port 18789 without authentication.

There are currently 923 Clawdbot instances running freely on the internet, according to this tweet: Tweet showing vulnerable instances

External risks

If the port is open without authentication: it's only a matter of time until someone takes control of your Clawdbot. Cybersecurity professionals have already warned about these vulnerabilities.

Don't slip up! Don't expose the port to the internet.

There are secure ways to access remotely: Tailscale (personal VPN) or Cloudflare Tunnel (zero-trust).

How do I know if I'm vulnerable?

Run the following command on the server running Clawdbot:

clawdbot security audit --deep

Validate the result: Auditing your Clawdbot security

To fix issues automatically: clawdbot doctor --fix

If you find something critical: fix it or shut it down immediately!

Useful commands

# Disable Clawdbot clawdbot gateway stop # Uninstall Clawdbot clawdbot gateway uninstall # Check Clawdbot status clawdbot gateway status

Integrations

When integrating with WhatsApp, Twitter and other social networks, every message becomes a prompt for the AI model (which has access to your terminal and tools), making it a potential attack vector.

WhatsApp

By default, the model:

  • Automatically reacts with 👀 to every mention or reply in groups
  • Automatically responds to mentions in groups (mode: mention)
  • Uses pairing for DMs: strangers receive a code and need to be approved

Disable automatic reaction:

clawdbot config unset messages.ackReaction

Change activation mode in groups:

# Only responds when mentioned (default) /activation mention # Responds to all group messages (careful!) /activation always

How does Pairing work?

When a stranger sends a DM to your Clawdbot:

  1. The bot doesn't process the message
  2. Sends a pairing code (e.g., A7X9)
  3. You approve manually:
# View pending requests clawdbot pairing list --channel whatsapp # Approve a contact clawdbot pairing approve whatsapp A7X9

This prevents anyone from using your bot without authorization.

Configure group policy

By default, the bot can respond in any group you're in. To restrict:

# Only responds in approved groups (recommended) clawdbot config set channels.whatsapp.groupPolicy "allowlist"

Why not start with WhatsApp?

  1. Confusing identity: the bot responds as if it were you (no visual distinction)
  2. Leak risk: any group mention becomes a prompt for the AI
  3. Attack surface: public groups = easier prompt injection

Prefer starting with Telegram, where the bot has its own identity and the experience is safer.

Disable read receipts

clawdbot config set channels.whatsapp.sendReadReceipts false

Internal risks

With great power comes great responsibility. Benjamin Franklin Parker

When installing it on your computer, you give unrestricted access to the model to all your files + terminal.

Clawdbot supports sandboxing via Docker to isolate sessions from untrusted groups and users.

Best practices:

  • Don't give Clawdbot write permissions on integrations you only need to read
  • Use sandboxing via Docker to isolate it from your computer.

Automatic integrations

Clawdbot connects to dozens of tools out-of-the-box:

Clawdbot vs Claude Cowork

Anthropic launched Claude Cowork in January 2026, currently has few integrations (files, google email, calendar) but I think it will evolve to become Anthropic's official version of Clawdbot.

Cheatsheet

Useful Clawdbot CLI Commands

Installation

npm install -g clawdbot@latest # Install clawdbot onboard --install-daemon # Setup daemon

Gateway

clawdbot gateway # Start gateway clawdbot gateway status # Check status clawdbot gateway stop # Stop clawdbot gateway uninstall # Uninstall clawdbot health # Health check clawdbot logs # View logs # Web UI: http://localhost:18789/

Security

clawdbot security audit # Basic audit clawdbot security audit --deep # Audit + network scan clawdbot security audit --fix # Apply auto-hardening

Diagnostics

clawdbot doctor # Diagnose issues clawdbot doctor --fix # Auto-fix problems clawdbot doctor --generate-gateway-token # Generate new token clawdbot status --all # Full status (safe to share)

Configuration

clawdbot config set <key> <value> # Set config clawdbot config unset <key> # Remove config clawdbot config get <key> # Get config

WhatsApp

# Group policy (allowlist = approved groups only) clawdbot config set channels.whatsapp.groupPolicy "allowlist" # Disable read receipts clawdbot config set channels.whatsapp.sendReadReceipts false # Disable auto-reaction (👀) clawdbot config unset messages.ackReaction

Pairing (contact approval)

clawdbot pairing list --channel whatsapp # List pending clawdbot pairing approve whatsapp <CODE> # Approve clawdbot pairing reject whatsapp <CODE> # Reject

Plugins

clawdbot plugins list # List plugins clawdbot plugins install <plugin> # Install plugin

In-chat commands

/activation mention # Only respond when mentioned /activation always # Respond to everything (careful!) /help # Show available commands

References

Official:

Security:

Tutorials:

Related:

Questions? @ocodista