Build Your Own Always-On AI Assistant with Claude, AWS Free Tier, and Telegram

Category: AI | Published: March 4, 2026

---

Introduction

What if your personal AI assistant never slept, never lost context, and lived entirely on infrastructure you control? That is no longer a hypothetical. By combining Clawd (an open-source Claude API runtime), a free AWS EC2 instance or any cheap VPS, and a messaging platform like Telegram, Discord, or WhatsApp, you can deploy a persistent, privacy-respecting AI assistant that is always reachable from your phone.

This guide walks through the full setup: provisioning a cloud server, installing Clawd, configuring your AI provider and model, and wiring everything up to a chat platform. Whether you are a developer who wants a smarter on-call helper or an automation engineer looking for a self-hosted alternative to SaaS AI wrappers, this stack is worth your attention.

---

Why Self-Host Claude Instead of Using a Managed Service?

Before jumping into the setup, it is worth understanding what you actually gain by running this yourself.

Data stays local. Clawd stores conversation history, memory, and context on the machine you control. No third-party SaaS layer is sitting between your prompts and the model API.

Always online. A cloud VM runs 24/7. Unlike a desktop app or a local LLM that requires your laptop to be awake, a server-based assistant receives messages, queues tasks, and sends proactive notifications even when you are offline.

Platform flexibility. Clawd supports multiple messaging connectors. You can talk to your assistant through Telegram's clean mobile UI, a private Discord server, or WhatsApp — whichever fits your existing workflow.

Cost. If you have never used AWS before, the AWS Free Tier gives you 750 hours per month of a t2.micro or t3.micro EC2 instance for the first 12 months. That is effectively a free always-on server to get started.

---

Step 1: Provision Your Server

Option A — AWS Free Tier (Recommended for Beginners)

1. Create an AWS account at aws.amazon.com if you do not have one.
2. Navigate to EC2 → Launch Instance.
3. Choose Ubuntu Server 24.04 LTS (free tier eligible).
4. Select the t2.micro instance type (1 vCPU, 1 GB RAM — sufficient for Clawd).
5. Create or reuse an SSH key pair and download the .pem file.
6. Under Security Group, open port 22 (SSH) and any port your messaging webhook will use (e.g., 8080 or 443).
7. Launch the instance and note the public IP address.

Option B — Existing VPS

If you already have a VPS with a provider like Hetzner, DigitalOcean, or Vultr — as the original author does — skip straight to the software setup. A VPS typically offers better network routing than a default AWS region, which matters if you need reliable access from regions with variable connectivity.

---

Step 2: Install and Configure Clawd

SSH into your server and run the following to get Clawd installed:

ssh -i your-key.pem ubuntu@your-server-ip

# Update system packages
sudo apt update && sudo apt upgrade -y

# Install Node.js (LTS) via nvm
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.7/install.sh | bash
source ~/.bashrc
nvm install --lts

# Install Clawd globally
npm install -g clawd

Once installed, initialize the configuration:

clawd init

This creates a config file (typically at ~/.clawd/config.yml) where you set your provider and model. For the Anthropic Claude API:

provider:
  name: anthropic
  api_key: YOUR_ANTHROPIC_API_KEY

model:
  default: claude-3-5-sonnet-20241022
  max_tokens: 4096

memory:
  backend: local
  path: ~/.clawd/memory

The memory.backend: local setting is the key privacy feature — all conversation history is written to your server's disk, not to any external service.

---

Step 3: Connect a Messaging Platform

This is where the setup moves from "useful" to "genuinely powerful." Clawd supports connectors for Telegram, Discord, and WhatsApp. Below are the essentials for the two most popular choices.

Telegram

1. Message @BotFather on Telegram and create a new bot. Copy the bot token.
2. Add the Telegram connector to your Clawd config:

connectors:
  telegram:
    enabled: true
    bot_token: YOUR_TELEGRAM_BOT_TOKEN
    allowed_users:
      - your_telegram_user_id

3. Start Clawd:

clawd start

Send your bot a message in Telegram. The assistant responds with full Claude-backed reasoning, maintains memory across sessions, and can proactively message you using the notifications API.

Discord

1. Create a Discord application and bot at discord.com/developers.
2. Invite the bot to a private server with the Send Messages and Read Message History permissions.
3. Add the Discord connector:

connectors:
  discord:
    enabled: true
    bot_token: YOUR_DISCORD_BOT_TOKEN
    channel_id: YOUR_CHANNEL_ID

Discord's threading model works particularly well if you want to separate different project contexts into different channels — each one acting as a distinct memory scope for the assistant.

---

Practical Use Cases

Once the stack is running, the real value reveals itself quickly. Here are patterns that work well in practice:

• Morning briefings. Schedule a cron job that triggers your assistant every morning to summarize your task list, pull from an RSS feed, or check a status API, then send you a Telegram message.
• On-call helper. Have your CI/CD pipeline send alert data to your assistant via a webhook. The assistant interprets the error, suggests a fix, and notifies you on Discord.
• Research assistant. Send a URL or a block of text from your phone. The assistant summarizes, extracts action items, and stores notes in local memory for future reference.
• Reminder system. Ask the assistant to remind you of something at a specific time. Because Clawd runs continuously, it can push a Telegram notification without any dependency on your local machine.

# Example: simple cron-triggered morning brief
0 8 * * * clawd run --task "Send morning summary to telegram"

---

Keeping the Service Running

To ensure Clawd restarts automatically after a server reboot, configure it as a systemd service:

# /etc/systemd/system/clawd.service
[Unit]
Description=Clawd AI Assistant
After=network.target

[Service]
ExecStart=/home/ubuntu/.nvm/versions/node/v22.0.0/bin/clawd start
Restart=always
User=ubuntu
Environment=NODE_ENV=production

[Install]
WantedBy=multi-user.target

sudo systemctl enable clawd
sudo systemctl start clawd

---

Conclusion

The combination of a free or low-cost cloud server, Clawd, and a messaging platform removes the two biggest limitations of most AI assistants: availability and data ownership. Your assistant runs continuously, remembers context across conversations, and delivers proactive notifications — all without your data passing through a third-party service you do not control.

The AWS Free Tier makes this accessible to anyone who wants to experiment without upfront cost. If you already run a VPS, the network and operational advantages make it an even cleaner choice.

Self-hosted AI tooling is maturing fast. This kind of setup — previously the territory of infrastructure-heavy teams — now takes an afternoon to configure. If you build something interesting on top of it, the ClawList community is the right place to share it.

---

Inspired by a tip from @kevinma_dev_zh. Tags: clawd claude telegram-bot discord-bot aws-free-tier self-hosted-ai vps ai-assistant