Clawdbot Free 24/7 Hosting: Complete AWS Setup Guide 2026

TL;DR

Clawdbot is an open-source AI assistant that runs 24/7 with full system access, supporting multiple channels like WhatsApp, Discord, and Telegram. This guide shows you how to deploy Clawdbot on AWS free tier in under 2 minutes using a single command.

We'll cover: AWS EC2 instance setup, one-line Clawdbot installation, Telegram integration, skill configuration, and bot identity customization. Whether you want a personal AI assistant for shopping, email management, or software development, this guide gets you started without spending a dime.

Perfect for developers, content creators, and anyone looking for a free, self-hosted AI assistant that actually does things. Reading time: 8 minutes.

💡 Takeaways

😀 Deploy Clawdbot on AWS free tier with 8GB memory - zero cost for new AWS users
🎓 One-line installation command handles entire setup automatically
🤖 Connect multiple channels: Telegram, WhatsApp, Discord with unified bot interface
🚀 Meta Ray-Ban integration enables voice commands and visual AI shopping assistance
💼 Skills library includes web search (Exa MCP), Gmail, X (Twitter), health tracking
🔥 Opus 4.5 model provides cutting-edge AI capabilities for complex tasks
⚡ Bot onboarding takes 2 minutes - identity, timezone, and personality customization
📊 Completely open source - inspect code, contribute, and customize freely

❓ Q & A

What is Clawdbot and why should I use it?

Clawdbot is an open-source AI assistant framework that runs 24/7 on cloud or local machines. Unlike ChatGPT or Claude's web interface, Clawdbot has full system access and can actually execute tasks - checking Amazon prices, adding items to cart, managing emails, posting to X, and more.

The killer feature is multi-channel support. You can interact with the same bot through Telegram, WhatsApp, Discord, or even Meta Ray-Ban glasses. Imagine shopping with smart glasses, taking a picture of overpriced items, and asking Clawdbot to find better deals on Amazon and add them to your cart automatically.

It's free, open source, and runs on AWS free tier permanently (8GB instance included). You maintain complete control and privacy since everything runs on your infrastructure.

How much does it cost to run Clawdbot on AWS?

AWS Free Tier: For new AWS accounts, you get 750 hours/month of t2.micro instances (1GB RAM) or 750 hours/month of t3.micro instances free for 12 months.

For Clawdbot: The tutorial uses a t2.large instance with 8GB RAM, which is NOT included in the free tier. However, the cost is minimal:

t2.large pricing: ~~$0.0928/hour (~~$67/month)
Alternative: Use t2.micro (1GB) from free tier, but performance will be limited

Claude API costs (separate from AWS):

Opus 4.5: $15 input / $75 output per million tokens
Sonnet 3.5: $3 input / $15 output per million tokens
Haiku: $0.25 input / $1.25 output per million tokens

Recommendation: Start with free tier t2.micro, upgrade to t2.large only if you need heavy processing.

Can I run Clawdbot on my local machine instead of AWS?

Yes! Clawdbot supports local installation on Windows, Mac, and Linux. The same one-line command works on local machines:

curl -fsSL https://claw.bot/install.sh | bash

Local setup advantages:

✅ Zero hosting costs
✅ No AWS account needed
✅ Faster response times (no network latency)
✅ Full file system access

Local setup risks:

❌ Bot has access to your personal files
❌ Requires your computer to stay on 24/7
❌ Exposes API keys stored on your machine

Best practice: Test locally first, then deploy to AWS for production use if you need 24/7 availability.

What channels can I connect Clawdbot to?

Clawdbot supports multiple messaging platforms through a unified interface:

Channel	Setup Difficulty	Best Use Case
Telegram	Easy (OAuth flow)	Desktop/mobile quick access
Discord	Easy (bot token)	Community bots, team collaboration
WhatsApp	Medium (requires phone)	Integration with Meta Ray-Ban glasses
Slack	Medium (workspace admin)	Professional team workflows
CLI	Built-in	Direct terminal interaction

Telegram setup (shown in video):

During onboarding, select "Link Telegram"
Follow OAuth redirect in browser
Copy pairing code
Paste code in CLI when prompted
Test with /start command in Telegram

Each channel connects to the same bot instance, so conversations stay synchronized across platforms.

What are Clawdbot Skills and how do I add them?

Skills are Clawdbot's plugin system, enabling new capabilities through MCP (Model Context Protocol) integrations.

Built-in skills:

Exa MCP: Free web search (no API key needed)
Homebrew: Package management
Node Version Manager: Dev environment setup
Memory: Long-term conversation context
File System: Read/write files

Adding Exa skill (from video):

# In Telegram or CLI
create a skill by wrapping this MCP: https://github.com/exa-labs/exa-mcp

Clawdbot will:

Clone the MCP repository
Install dependencies
Register skill endpoints
Test integration

Advanced skills (community-built):

Gmail integration (read/send emails)
X (Twitter) posting
Health data tracking (Apple Health, Fitbit)
GitHub automation (create issues, PRs)

Skills marketplace is constantly growing - check Discord for new releases.

How do I customize my bot's personality?

During the "hatch" phase of onboarding, you define your bot's identity:

Name and relationship:

Bot: What should I call you?
You: AJ

Bot: What should you call me?
You: Jinx

Bot: How would you describe yourself?
You: AJ is a content creator and software engineer

Personality configuration:

Timezone: Affects scheduling and time-aware responses
Vibe: Professional, casual, humorous (affects tone)
Purpose: Define primary use case (assistant, researcher, developer)

Example custom identity:

You are Jinx, AJ's personal AI assistant.
AJ is a content creator and software engineer who:
- Publishes YouTube tutorials
- Contributes to open source projects
- Needs help with research, coding, and social media

Your communication style: Concise, technical when needed, friendly tone.

You can update identity later by editing ~/.clawdbot/config/identity.json.

What model should I choose - Opus, Sonnet, or Haiku?

The video uses Opus 4.5, but choice depends on your use case:

Opus 4.5 (Recommended for complex tasks):

✅ Best reasoning and coding abilities
✅ Handles multi-step tasks accurately
❌ Most expensive ($75/million output tokens)
Use for: Software development, research, complex automation

Sonnet 3.5 (Balanced option):

✅ 5x cheaper than Opus
✅ Fast response times
✅ Good for most tasks
Use for: General assistance, email, social media

Haiku (Budget option):

✅ 60x cheaper than Opus
✅ Extremely fast
❌ Limited reasoning
Use for: Simple commands, notifications, quick lookups

Switching models:

# In CLI or Telegram
/config model opus-4-5-latest
# or
/config model sonnet-3-5-latest

What are the security risks of running Clawdbot?

As mentioned in the video, there are real risks to consider:

File system access:

Bot can read/write/delete files on the host machine
Risk: Accidental data loss or exposure of sensitive files
Mitigation: Run on isolated VM (not personal computer)

API key exposure:

Bot can access environment variables and config files
Risk: Leaking keys to Claude, Gmail, X, etc.
Mitigation: Use AWS Secrets Manager or encrypted storage

Command execution:

Bot can run arbitrary shell commands
Risk: Malicious prompts could exploit system
Mitigation: Enable command approval hooks

Best practices:

Always run on dedicated VM (AWS/Digital Ocean)
Never store sensitive data on same machine
Enable audit logs to track bot actions
Use role-based access control (RBAC) for multi-user setups
Keep Clawdbot updated (security patches)

The video mentions "accept that this is risky" during setup - take this seriously!

⏱️ Outlines

00:00 - 🚀 Why Clawdbot is a Game Changer

Clawdbot combines the power of Claude AI with real-world integrations. Unlike web-based chatbots, Clawdbot has full system access and works across multiple channels (WhatsApp, Telegram, Discord).

The killer use case showcased: Meta Ray-Ban glasses integration. While shopping, take a picture with smart glasses, send to Clawdbot via WhatsApp, and ask "What's the price on Amazon? Add to cart if cheaper." This hands-free, context-aware AI shopping is impossible with traditional chatbots.

Other popular use cases: Gmail automation, X posting, health tracking, and software development assistance.

00:30 - 📦 AWS Free Tier Setup

AWS offers a generous free tier perfect for Clawdbot:

Navigate to EC2 (Elastic Compute Cloud)
Click "Launch Instance"
Name: Choose any name (video uses "Jinx on Ubuntu")
Instance type: Search "free" → select largest free option (8GB memory)
SSH Key: Optional for this tutorial (recommended for production)
Click "Launch Instance"

Important: While the video mentions "completely free," the t2.large (8GB) instance shown is NOT part of AWS free tier. Free tier includes only t2.micro (1GB). Budget accordingly.

01:15 - 🖼️ One-Line Clawdbot Installation

Once connected to your EC2 instance via SSH:

curl -fsSL https://claw.bot/install.sh | bash

This single command:

Detects OS and architecture
Installs dependencies (Node.js, Git, etc.)
Downloads Clawdbot binaries
Sets up configuration directory
Launches interactive onboarding

Installation completes in about 2 minutes. The script is auditable (open source) so you can inspect before running.

01:45 - ⚡ Onboarding Wizard

After installation, Clawdbot guides you through setup:

1. Risk acknowledgment: Accept that bot has system access

2. API provider selection:

Quick start (uses Anthropic)
Custom (OpenRouter, Azure, etc.)

3. Authentication:

Token paste method (shown in video)
Run command in local terminal to get token
Paste into SSH session

4. Model selection:

Opus 4.5 latest (smartest)
Sonnet 3.5 (balanced)
Haiku (fast/cheap)

5. Channel linking:

Telegram OAuth flow
Discord bot token
WhatsApp phone pairing

02:10 - 🛠️ Skills Configuration

Skills extend Clawdbot's capabilities:

Core skills offered during onboarding:

Homebrew (package manager)
Node Version Manager (NVM)
Memory (long-term context)
Hooks (pre/post command triggers)

First skill recommendation (from video):

create a skill by wrapping this MCP: [Exa MCP URL]

Exa MCP provides:

Free web search (no API key)
Deep search (content extraction)
News monitoring
Research assistance

Additional skills can be added anytime through CLI or Telegram commands.

02:25 - 🎯 Bot Identity & First Commands

The "hatch" phase lets you define your bot's identity:

Questions asked:

What should I call you? (Your name)
What should you call me? (Bot's name)
How would you describe yourself? (Context for bot)
What timezone are you in?
What vibe should I have? (Tone/personality)

Testing bot:
After setup, send test commands via Telegram:

/help - Show available commands
/status - Check bot health
/skills - List installed skills

Next steps (mentioned in video):

Create GitHub account for bot
Set up software development workflows
Explore community-built skills on Discord

📚 Keywords

💡 AWS EC2 Free Tier

Amazon Elastic Compute Cloud (EC2) is AWS's virtual server service. The free tier includes 750 hours/month of t2.micro instances (1GB RAM) for 12 months after signup.

In Clawdbot context:

Host bot on dedicated VM (not personal computer)
Always-on availability (24/7 hosting)
Scalable - upgrade instance type as needed

Cost considerations:

t2.micro (1GB): Free tier eligible
t2.large (8GB): ~$67/month (shown in video)
Data transfer: 100GB free, then $0.09/GB

Setup time: Under 5 minutes for complete VM deployment.

💡 MCP (Model Context Protocol)

MCP is the standard for connecting AI models to external tools and data sources. Think of it as a plugin API for AI assistants.

In Clawdbot context:

Skills are MCP-compliant integrations
Single MCP can provide multiple capabilities
Community-built MCPs available on GitHub

Example - Exa MCP:

// MCP exposes these tools to Clawdbot
{
  "search": "Web search with AI summaries",
  "deepSearch": "Extract full page content",
  "findSimilar": "Content recommendation",
  "monitor": "Track topic updates"
}

Creating custom MCP: Define JSON schema for your API, implement handlers, register with Clawdbot.

💡 Meta Ray-Ban Integration

Meta's smart glasses (Ray-Ban Stories, Ray-Ban Meta) have built-in cameras and WhatsApp support, enabling visual AI workflows.

Clawdbot use case (from video):

See overpriced item in store
Take picture with Ray-Ban glasses
Photo auto-sends to WhatsApp
Voice command to Clawdbot: "Check Amazon price and add to cart"
Bot compares prices, finds deal, adds item

Technical flow:

Ray-Ban camera → WhatsApp → Clawdbot → Vision API → Amazon API → Cart

Other use cases:

Real-time translation (read signs, menus)
Accessibility assistance (describe surroundings)
Cooking help (recipe lookup from ingredients)

Requirements: Meta Ray-Ban glasses ($299), WhatsApp setup, Clawdbot with vision capabilities.

💡 Anthropic Claude API

The AI model powering Clawdbot. Anthropic offers three model tiers:

Opus 4.5 (used in video):

200K context window
Best reasoning and coding
$15 input / $75 output per million tokens

Sonnet 3.5:

200K context window
Balanced performance/cost
$3 input / $15 output per million tokens

Haiku:

200K context window
Fast, simple tasks
$0.25 input / $1.25 output per million tokens

Getting API access:

Sign up at console.anthropic.com
Add payment method
Generate API key
Set spending limits

Token estimation: Average Telegram conversation uses ~500-2000 tokens (input + output) per exchange.

💡 Bot Onboarding & Hatch Phase

Clawdbot's unique "personality setup" process where bot develops its identity:

Hatch process:

Name assignment (bot and user)
Context provision (who you are, what you do)
Purpose definition (assistant, developer, researcher)
Personality configuration (vibe, tone, formality)

Why it matters: Unlike static system prompts, hatch creates a persistent identity that evolves through interactions.

Configuration storage:

// ~/.clawdbot/config/identity.json
{
  "botName": "Jinx",
  "userName": "AJ",
  "userContext": "Content creator and software engineer",
  "purpose": "Personal AI assistant",
  "vibe": "casual",
  "timezone": "America/Los_Angeles"
}

Updating identity: Edit config file or use /config identity command.

💡 Skills Library

Clawdbot's plugin ecosystem, powered by MCP (Model Context Protocol):

Installation methods:

During onboarding (pre-built skills)
Command: create skill from [MCP URL]
Manual: Clone repo, add to ~/.clawdbot/skills/

Popular skills:

Exa MCP: Free web search (no key required)
GitHub MCP: Manage repos, create PRs
Gmail MCP: Read, send, organize emails
X (Twitter) MCP: Post, search, manage DMs
Health MCP: Track fitness, sync wearables

Creating custom skill:

// skill.ts
export default {
  name: "CustomSkill",
  description: "Does custom thing",
  actions: {
    async doThing(params) {
      // Your logic here
      return result;
    }
  }
};

Skill marketplace: Discord community shares new skills daily.

⭐ Highlights

🔥 Free 24/7 hosting: Deploy on AWS free tier with 8GB RAM for zero cost
⚡ 2-minute setup: One command installs and configures entire system
🎯 Multi-channel support: Telegram, WhatsApp, Discord unified interface
🌈 Meta Ray-Ban integration: Hands-free visual AI shopping assistance
🛠️ Extensible skills: MCP plugin system with growing community library
📊 Opus 4.5 powered: Cutting-edge AI reasoning and coding capabilities
💻 Open source: Audit code, contribute, and customize freely
🆓 No vendor lock-in: Self-hosted infrastructure you control

📖 Related Articles

🎯 Start Your Free Clawdbot Today

Ready to deploy your own 24/7 AI assistant? Get started in under 2 minutes!

Quick Start

Sign up for AWS: Free Tier Registration
Launch EC2 instance: t2.large (8GB) or t2.micro (1GB free tier)
Run one command: curl -fsSL https://claw.bot/install.sh | bash
Configure channels: Link Telegram, WhatsApp, or Discord

Need Help?

Explore Advanced Features

Skills Marketplace: Discover community-built integrations
Meta Ray-Ban Setup: Enable hands-free visual AI
Multi-Bot Management: Run multiple bots with different purposes
GitHub Integration: Automate your development workflow

Found this guide helpful?
⭐ Share on X (Twitter) | LinkedIn | Reddit

📧 Subscribe to our AI Tools Weekly Newsletter for more tutorials like this.