RR

Getting Started with OpenClaw

OpenClaw section · 12 min read · Last updated February 2026

This guide covers everything you need to go from zero to a working OpenClaw agent — the hardware, the accounts, the installation, and the first configuration that makes it actually useful. No fluff, no assumed knowledge.

On this page
  1. What you'll have at the end
  2. Requirements
  3. Install OpenClaw
  4. Connect Telegram
  5. Connect an AI model
  6. First test
  7. Configure your agent's identity
  8. What to do next

1. What You'll Have at the End

A working agent that:

This is the foundation. Everything else — cron schedules, additional tools, memory architecture, specialized skills — gets layered on top. Get this working first.

2. Requirements

Hardware
  • Mac (Apple Silicon recommended)
  • 8GB RAM minimum, 16GB+ preferred
  • macOS 13 Ventura or newer
  • Stable internet connection
Accounts
  • Telegram account + phone number
  • Anthropic account (or OpenAI)
  • Brave Search API key (free tier works)
Software
  • Node.js v18 or newer
  • Homebrew (macOS package manager)
  • Git
Cost to get started
  • OpenClaw: free, open source
  • Telegram bot: free
  • Claude API: pay-per-use (~$5–20/mo typical)
  • Brave Search: 2,000 free queries/mo
Apple Silicon recommendation: An M4 Mac mini (16GB, $599) is the best entry point. The unified memory architecture means it can also run local models later — giving you a path to zero API costs as you scale. You don't need it to start, but it's the right long-term platform.

3. Install OpenClaw

Install Homebrew if you don't already have it:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

Install Node.js:

brew install node

Install OpenClaw globally via npm:

npm install -g openclaw

Verify the install:

openclaw --version

Initialize your workspace:

openclaw init

This creates a workspace directory at ~/.openclaw/workspace — the folder your agent reads from, writes to, and calls home.

4. Connect Telegram

Create a Telegram bot:

  1. Open Telegram and search for @BotFather
  2. Send /newbot
  3. Choose a name and username for your bot
  4. BotFather will give you an API token — copy it

Add the token to your OpenClaw config:

openclaw config set telegram.botToken YOUR_TOKEN_HERE

Get your Telegram user ID:

  1. Search for @userinfobot on Telegram
  2. Send it any message — it replies with your numeric ID
openclaw config set telegram.allowedUsers YOUR_TELEGRAM_ID
Why this matters: The allowedUsers setting locks your bot so only you (and whoever else you explicitly add) can talk to it. Without this, anyone who finds your bot can send it commands.

5. Connect an AI Model

OpenClaw supports multiple model providers. For a first setup, Anthropic Claude is the most capable and the most consistent.

Get an API key from console.anthropic.com. Add it:

openclaw config set anthropic.apiKey YOUR_KEY_HERE
openclaw config set model anthropic/claude-sonnet-4-6

If you want to run a local model instead (no API costs):

# Install Ollama
brew install ollama

# Pull a model (16GB RAM minimum for 7B models)
ollama pull llama3.2

# Point OpenClaw at it
openclaw config set model ollama/llama3.2
Which model to use: Start with Claude if you have an API key. It's the most capable out of the box and handles complex instructions well. Switch to a local model once you understand what tasks your agent is actually running — local models are free and handle most recurring jobs fine.

6. First Test

Start the OpenClaw gateway:

openclaw gateway start

Open Telegram, find your bot, and send it a message:

Hello — what can you do?

If it responds, you have a working agent. If it doesn't:

  • Check gateway status: openclaw gateway status
  • Check logs: openclaw logs
  • Verify your bot token is correct in the config
  • Confirm you messaged the right bot (search by username)

Once it's responding, test a basic task:

Search the web for the latest news on AI agents and give me a 3-bullet summary.

If that works, your agent has a model, memory, and web access. The core is running.

To have the gateway start automatically when you restart your Mac:

openclaw gateway enable

7. Configure Your Agent's Identity

This is the most important configuration step and the one most people skip. Your agent's identity file is what separates a reliable system from one that drifts.

Open your workspace:

cd ~/.openclaw/workspace

You'll find a SOUL.md file. This is where you define:

  • What your agent's name and role is
  • What it will never do (hard limits)
  • How it communicates
  • What it should do proactively vs. only when asked

There's also AGENTS.md (operating rules), MEMORY.md (long-term memory), and a memory/ folder. These files are what the agent reads at the start of every session to understand who it is and what it's working on.

Don't skip this. An agent without a configured identity is just a chat interface. The identity files are what make it an operator — something that acts consistently, remembers context, and behaves predictably over time.

8. What to Do Next

Once you have a working agent, the next questions are:

These questions are where the real setup work happens. The installation above gets you a running system. The configuration work is what makes it useful.

The full configuration guide — identity setup, memory architecture, cron automation, Telegram integration, and deployment — is covered step by step in the AI Ops Setup Guide. Everything I actually run, documented from scratch.

Want to see what day-to-day looks like as you build this out? Follow the Daily Log →

Want a guided walkthrough + coaching on implementation?

We'll walk through your setup, your workflows, and get it running the right way.

Book a Call →

© Ridley Research & Consulting. All rights reserved.