How to Install OpenClaw and Connect Telegram: Step-by-Step Beginner Guide

Introduction

This is Day 2 of the 16-day OpenClaw Bootcamp. Today you go from zero to a fully working AI agent that runs on your own machine and responds to you through Telegram. By the end of this session, you'll have a live AI system you can message from your phone.

No prior AI experience required. If you can open a terminal and type a few commands, you can do this.

Video Tutorial

Watch the full Day 2 video with screen recordings of every step:

What You'll Build Today

By the end of Day 2, you'll have:

1. OpenClaw installed on your machine
2. The gateway running and accessible via the dashboard
3. An AI model connected (Anthropic, OpenAI, or another provider)
4. Telegram fully set up — both the bot and your personal account paired
5. A working conversation with your agent through Telegram

Install Node.js

Node.js is the only prerequisite. OpenClaw runs on Node, so you need it installed before anything else.

If you're not sure whether you have it, open your terminal and type node --version. If you see a version number (v18 or higher), you're good. If not, download the latest LTS version from the official Node.js website and run the installer.

That's it for prerequisites. No Docker, no Python, no complicated environment setup. Just Node.

Run the OpenClaw Installer

With Node installed, running the OpenClaw installer is a single command in your terminal. The installer downloads OpenClaw and launches the onboarding wizard — an interactive setup that walks you through:

• Creating your workspace directory
• Choosing your AI model provider (Anthropic, OpenAI, etc.)
• Entering your API key
• Setting your agent's basic personality via soul.md
• Configuring your first channel

The wizard handles all the file creation and config setup. You just answer questions.

Start the Gateway

After the installer finishes, start the OpenClaw gateway. This is the central process that keeps your agent running — it handles message routing, heartbeats, memory management, and channel connections.

Once the gateway starts, you'll see a URL for the dashboard. Open it in your browser to see your agent's status, recent messages, and configuration. The dashboard is your control center.

Send Your First Message

With the gateway running, send your first message through the dashboard's web chat. Type something simple like "Hello, what can you do?" and verify your agent responds. If you get a response, your system is working — the model is connected, the gateway is routing messages, and your agent is alive.

Understanding the Workspace & Config

Your OpenClaw workspace is a directory on your machine that contains everything your agent needs:

• soul.md — your agent's personality, instructions, and system prompt
• config files — model settings, channel configs, heartbeat intervals
• memory/ — your agent's long-term memory files
• skills/ — any custom skills you've added

Everything is just files on your disk. There's no database, no cloud backend. You can read, edit, backup, and version-control your entire agent setup.

Telegram Setup (Bot + Account Pairing)

Telegram is the most popular first channel for OpenClaw. Setup has two parts:

Part 1: Create a Telegram Bot

Open Telegram and message @BotFather. Create a new bot, give it a name, and you'll receive a bot token. This token goes into your OpenClaw config.

Part 2: Pair Your Personal Account

The bot lets your agent receive messages, but pairing your personal account lets your agent send proactive messages to you, access group chats, and behave more like a real contact than a bot.

Follow the pairing flow in the OpenClaw setup — it generates a QR code or pairing code that links your Telegram account to your agent.

Testing

Once both parts are connected, open Telegram and message your bot. You should see your agent respond in real-time. Try a few back-and-forth messages to verify everything is working. Congratulations — you now have an AI agent running on your machine that you can talk to from your phone.

Common Errors and Fixes

• "Cannot find module" error: Your Node.js version is too old. Update to v18 or higher.
• API key rejected: Double-check your key — make sure there are no extra spaces. Verify your account has API access enabled and billing set up.
• Gateway starts but no response: Check the dashboard logs. Usually a model configuration issue — the API key, model name, or provider might be misconfigured.
• Telegram bot doesn't respond: Verify the bot token is correct in your config. Make sure the gateway is running. Check that you're messaging the right bot (not @BotFather).
• Port already in use: Another process is using the gateway port. Either stop that process or change the port in your OpenClaw config.

Security Habits to Build Now

Start building these habits from Day 1:

• Set API spending limits — before leaving your agent running unattended, set a cap on your API provider. $20–50/month is a safe starting limit.
• Use allowlists — restrict which Telegram users can talk to your agent. Don't leave it open to the public.
• Keep API keys out of git — if you version-control your workspace, use environment variables or a .env file for sensitive data.
• Review the dashboard regularly — check who's sending messages and how much it's costing you.

What You Have Now + Day 3 Preview

At the end of Day 2, you have:

• A running OpenClaw agent on your local machine
• A connected AI model processing your messages
• Telegram set up for mobile access
• A web dashboard for monitoring

In Day 3, we dive into model selection — you'll compare Claude, GPT, Gemini, and local models, and connect an Ollama model so your agent can run for $0 on your own hardware.

Frequently Asked Questions

Do I need a powerful computer?

No. For cloud models (Claude, GPT), OpenClaw itself is very lightweight — it's just a Node.js process. Any modern laptop works fine. You only need serious hardware if you want to run local models (covered in Day 3).

Can I install OpenClaw on Windows?

Yes. OpenClaw runs on macOS, Linux, and Windows. The installer and setup process is the same on all platforms.

How much will the API cost me?

For light personal use (a few conversations per day), expect $5–15/month with Claude Sonnet or GPT-4o. We cover cost optimization in depth on Day 4, where most users get their costs down to $15–20/month even with heavy use.

Can I skip Telegram and use a different channel?

Yes, but we recommend starting with Telegram because it's the fastest to set up and has the best feature support. You can add other channels (WhatsApp, Discord, Slack) on Day 5.

What if I don't have an API key yet?

Sign up for an Anthropic or OpenAI account and generate an API key. Both offer free trial credits for new accounts. You can also use Google Gemini's free tier to start with zero cost.