AI ASSISTANT

Clawdbot Setup Guide: Your Personal AI Assistant for Any Platform 🦞

Clawdbot is a personal AI assistant you run on your own devices. It answers you on WhatsApp, Telegram, Slack, Discord, Signal, iMessage, and more—with voice support, live Canvas, and full control over your data.

January 26, 2026
12 min read
AI, Messaging, Automation

What is Clawdbot?

Clawdbot is a personal AI assistant that you run on your own devices. Unlike cloud-based AI services, Clawdbot gives you full control—it connects to the messaging channels you already use (WhatsApp, Telegram, Slack, Discord, and more) and responds as your personal AI assistant.

Built by Peter Steinberger and the open-source community, Clawdbot supports multiple AI providers including Anthropic (Claude) and OpenAI. The project has over 31,000 stars on GitHub and is actively maintained with regular releases.

🦞 What Makes Clawdbot Special
  • Multi-Channel: Works on WhatsApp, Telegram, Slack, Discord, Signal, iMessage, Microsoft Teams, Google Chat, and more
  • Self-Hosted: Run on your own devices—your data stays with you
  • Voice Support: Voice Wake and Talk Mode for hands-free interaction
  • Live Canvas: Agent-driven visual workspace for rich interactions
  • Multi-Agent Routing: Route different channels to isolated agents
  • Companion Apps: Native macOS menu bar app, iOS and Android nodes

Why Choose Clawdbot?

Own Your Data

Unlike cloud AI services, Clawdbot runs on your hardware. Your conversations, context, and data never leave your control.

  • Self-hosted on your devices
  • No third-party data storage
  • Full privacy control
  • Local-first architecture
Use Your Channels

Chat with your AI assistant on the platforms you already use daily. No new apps to install—just message your assistant.

  • WhatsApp, Telegram, Discord
  • Slack, Microsoft Teams
  • Signal, iMessage (macOS)
  • Google Chat, Matrix, WebChat
💡 Perfect For

Personal Use

  • Personal AI assistant on your phone
  • Quick Q&A via messaging apps
  • Voice-activated tasks
  • Always-on availability

Power Users

  • Multi-agent workflows
  • Custom skills and tools
  • Browser automation
  • Cron jobs and webhooks

Prerequisites & Requirements

System Requirements
  • Node.js: Version 22 or higher (required)
  • OS: macOS, Linux, or Windows (WSL2 strongly recommended)
  • pnpm: Recommended if building from source
  • macOS: Xcode/CLT if building companion apps
AI Provider Auth
  • Anthropic: API key or Claude Pro/Max subscription (OAuth)
  • OpenAI: ChatGPT/Codex subscription (OAuth) or API key
  • Recommended: Anthropic Pro/Max + Opus 4.5 for best results
  • Optional: Brave Search API key for web search
⚠️ Windows Users

WSL2 (Ubuntu recommended) is strongly recommended for Windows. Native Windows is untested and has poorer tool compatibility. Install WSL2 first, then run all commands inside WSL.

Installation Guide

The recommended approach is to use the CLI onboarding wizard, which sets up everything automatically.

Step 1: Install the CLI

Install via the official installer script:

bashterminal
# macOS / Linux
curl -fsSL https://clawd.bot/install.sh | bash

# Windows (PowerShell) - then use WSL2
iwr -useb https://clawd.bot/install.ps1 | iex

# Alternative: npm/pnpm global install
npm install -g clawdbot@latest
# or
pnpm add -g clawdbot@latest

Step 2: Run the Onboarding Wizard

The wizard guides you through gateway setup, auth, channels, and daemon installation:

bashterminal
clawdbot onboard --install-daemon

The wizard will help you configure:

  • Local vs Remote gateway mode
  • Auth: OpenAI Codex subscription (OAuth) or API keys
  • Providers: WhatsApp QR login, Telegram/Discord tokens, etc.
  • Daemon: Background service (launchd/systemd)
  • Gateway token: Auto-generated for security

Step 3: Verify the Gateway

bashterminal
# Check gateway status
clawdbot gateway status

# Run gateway manually (foreground)
clawdbot gateway --port 18789 --verbose

# Quick health check
clawdbot status
clawdbot health
💡 Dashboard Access

Once the gateway is running, access the Control UI dashboard at http://127.0.0.1:18789/

Key Features & Capabilities

Multi-Channel Inbox
  • WhatsApp (via Baileys)
  • Telegram (via grammY)
  • Slack (via Bolt)
  • Discord (via discord.js)
  • Signal, iMessage, Microsoft Teams
  • Google Chat, Matrix, WebChat
Voice & Canvas
  • Voice Wake for always-on listening
  • Talk Mode for continuous conversation
  • Live Canvas visual workspace (A2UI)
  • ElevenLabs voice integration
  • macOS/iOS/Android support
Tools & Automation
  • Browser control (dedicated Chromium)
  • Camera snap/clip, screen record
  • Cron jobs and webhooks
  • Gmail Pub/Sub triggers
  • Skills platform (bundled + custom)
Companion Apps
  • macOS menu bar app
  • iOS node (Canvas, Voice, camera)
  • Android node (Canvas, Talk, camera)
  • Remote gateway control
🔌 Multi-Agent Routing

Route different channels, accounts, or peers to isolated agents with their own workspaces and sessions. Perfect for separating personal and work contexts.

jsonclawdbot.json
{
  "routing": {
    "agents": {
      "main": {
        "workspace": "~/clawd",
        "sandbox": { "mode": "off" }
      },
      "work": {
        "workspace": "~/clawd-work",
        "sandbox": { "mode": "non-main" }
      }
    }
  }
}

Supported Channels

WhatsApp

Link via QR code, just like WhatsApp Web.

bash
clawdbot channels login
Telegram

Use a bot token from @BotFather.

json
{
  "channels": {
    "telegram": {
      "botToken": "123:ABC"
    }
  }
}
Discord

Configure with your Discord bot token.

json
{
  "channels": {
    "discord": {
      "token": "your-token"
    }
  }
}
Slack

Use Bolt with bot and app tokens.

bash
SLACK_BOT_TOKEN=xoxb-...
SLACK_APP_TOKEN=xapp-...
Signal

Requires signal-cli setup.

iMessage

macOS only, requires Messages sign-in.

Getting Started: Your First Chat

1. Pair a Channel (WhatsApp Example)

bashterminal
# Login to WhatsApp (scan QR code)
clawdbot channels login

# Approve pairing for new contacts
clawdbot pairing list whatsapp
clawdbot pairing approve whatsapp <code>

2. Send a Test Message

bashterminal
# Send a message via CLI
clawdbot message send --to +1234567890 --message "Hello from Clawdbot"

# Or interact directly with the agent
clawdbot agent --message "What can you help me with?" --thinking high

3. Chat Commands

Send these commands in any connected channel (WhatsApp, Telegram, etc.):

CommandDescription
/statusShow session status (model, tokens, cost)
/new or /resetReset the session
/compactCompact session context (summary)
/think <level>Set thinking level (off/minimal/low/medium/high/xhigh)
/verbose on|offToggle verbose mode
/restartRestart gateway (owner-only)

Advanced Usage & Configuration

Configuration File

Main configuration lives at ~/.clawdbot/clawdbot.json:

json~/.clawdbot/clawdbot.json
{
  "agent": {
    "model": "anthropic/claude-opus-4-5"
  },
  "browser": {
    "enabled": true,
    "controlUrl": "http://127.0.0.1:18791"
  },
  "channels": {
    "whatsapp": {
      "allowFrom": ["+1234567890"]
    }
  }
}

Skills & Workspace

Clawdbot supports bundled, managed, and workspace skills:

  • Workspace root: ~/clawd (configurable)
  • Prompt files: AGENTS.md, SOUL.md, TOOLS.md
  • Custom skills: ~/clawd/skills/<skill>/SKILL.md

Tailscale Remote Access

Expose the gateway securely via Tailscale Serve or Funnel:

jsonclawdbot.json
{
  "gateway": {
    "tailscale": {
      "mode": "serve"  // or "funnel" for public access
    }
  }
}

Building from Source

bashterminal
git clone https://github.com/openclaw/openclaw.git
cd clawdbot
pnpm install
pnpm ui:build    # auto-installs UI deps
pnpm build
pnpm clawdbot onboard --install-daemon

# Development mode (auto-reload)
pnpm gateway:watch

Security & Best Practices

Important: DM Security

Clawdbot connects to real messaging surfaces. Treat all inbound DMs as untrusted input.

  • Default behavior: Unknown senders receive a pairing code— the bot won't process messages until approved
  • Approve senders: clawdbot pairing approve <channel> <code>
  • Run diagnostics: clawdbot doctor to check for risky DM policies
Security Defaults
  • DM pairing required by default
  • Tools run on host for main session only
  • Non-main sessions can be sandboxed (Docker)
  • Gateway tokens auto-generated
Sandbox Mode
  • Set sandbox.mode: "non-main"
  • Groups/channels run in Docker
  • Bash isolated per session
  • Tool allowlist/denylist supported
💡 Pro Tip

Run clawdbot doctor regularly to check for configuration issues and security risks. It will surface any misconfigured DM policies or missing auth.

Clawdbot Pricing: Is It Free?

Clawdbot itself is free and open source under the MIT license. You can clone the Clawdbot GitHub repository, build it, and run it without paying anything for the software.

However, Clawdbot uses Claude as its AI backend, so you need an Anthropic API key. That means you pay for the Claude API usage. The cost depends on which Claude model you configure and how much you chat. Typical personal usage runs $5-20 per month on the API. You can set spending limits in your Anthropic dashboard to control costs.

ComponentCostNotes
Clawdbot softwareFreeMIT license, open source
Anthropic API (Claude)Pay-per-use~$5-20/month for typical personal use
Hosting (self-hosted)FreeRun on your own Mac, Linux, or Raspberry Pi
Hosting (cloud VPS)$5-10/monthOptional, if you want 24/7 uptime without a home server

Bottom line: Clawdbot is free to install and use. You only pay for the underlying Claude API calls, which you control directly through your Anthropic account. There are no subscriptions, no premium tiers, and no usage limits imposed by Clawdbot itself.

Clawdbot vs Openclaw vs Moltbot

Clawdbot, Openclaw, and Moltbot are three related projects in the open-source AI assistant ecosystem. They share a common goal - giving you a personal AI that runs on your own hardware - but take different approaches to get there. Here is how they compare.

FeatureClawdbotOpenclawMoltbot
AI backendClaude (Anthropic)Local models (Ollama, MLX)Multi-provider (Claude, GPT, local)
Data privacyData goes to Anthropic APIFully local, no external callsDepends on provider chosen
Setup difficultyEasy (CLI wizard)Medium (model download + config)Medium
Messaging channelsWhatsApp, Telegram, Discord, Slack, Signal, iMessageTerminal, web UITelegram, Discord, web UI
QualityHigh (Claude-powered)Varies by local modelVaries by provider
Tool useYes (MCP, skills)LimitedYes
Running costAPI fees ($5-20/mo)Free (local compute)Free to API fees

Pick Clawdbot if you want the highest quality AI responses through Claude, the widest messaging app support (especially WhatsApp and iMessage), and the easiest setup with the CLI wizard. Clawdbot is the most polished option for non-technical users who want a personal AI assistant on their phone.

Pick Openclaw if data privacy is your priority and you do not want any data leaving your machine. Openclaw runs entirely on local models through Ollama or MLX, so nothing hits an external API. The tradeoff is lower output quality compared to Claude and fewer messaging integrations.

Pick Moltbot if you want flexibility to switch between providers. Moltbot lets you route different conversations to different AI backends - use Claude for important chats and a local model for casual ones. It is a good middle ground if you want both quality and privacy depending on the context.

Running Clawdbot on a Mac Mini

A Mac Mini is one of the best home server options for running Clawdbot 24/7. The M2 or M4 Mac Mini draws just 5-10 watts at idle, runs silently, and handles Clawdbot with plenty of headroom. Here is how to set it up as a dedicated Clawdbot server.

bashterminal
# Install Clawdbot on Mac Mini
brew install clawdbot/tap/clawdbot

# Run the onboarding wizard
clawdbot onboard

# Start Clawdbot as a background service (survives reboots)
brew services start clawdbot

# Check status
brew services list | grep clawdbot

# View logs
tail -f ~/Library/Logs/clawdbot/clawdbot.log
Mac Mini Hardware Recommendations
  • Mac Mini M2 (8GB): Runs Clawdbot comfortably. More than enough for the gateway, messaging bridges, and API calls. This is the most cost-effective option.
  • Mac Mini M2/M4 (16GB+): Can run Clawdbot plus a local model via Ollama simultaneously - useful if you want to experiment with Openclaw-style local inference alongside Clawdbot.
  • Power usage: 5-10W at idle. Your electricity cost is under $2 per month to run Clawdbot 24/7 on a Mac Mini.

For remote access when you are away from home, set up Tailscale on the Mac Mini. This lets you SSH into it and manage Clawdbot from anywhere without opening ports on your router. Clawdbot's messaging bridges handle the rest - your WhatsApp, Telegram, and Discord channels stay connected regardless of your location.

Frequently Asked Questions

Ready to Get Started?

Clawdbot is the personal AI assistant you run on your own terms. With support for 12+ messaging channels, voice interaction, and a thriving open-source community, it's the most flexible way to have an AI assistant everywhere you chat.

Related Articles

AI Integration
Jenkins MCP Server
Enable AI agents to interact with Jenkins CI/CD infrastructure.
MCP Server
Calculator Server
Go-based MCP server for mathematical computations.
CRM Integration
Method CRM MCP Server
Production-ready MCP server for Method CRM API integration.