Guide

The Complete Guide to OpenClaw Setup (2026)

Everything you need to install, configure, and deploy OpenClaw on your own hardware. Updated for 2026.

What is OpenClaw?

OpenClaw is an open-source AI assistant framework that runs on your own hardware. Unlike cloud-only solutions, OpenClaw gives you full control over your data, your models, and your integrations. It connects to Discord, Telegram, Slack, email, SMS, and more — acting as an always-on AI agent for your business or personal workflows.

OpenClaw supports multi-agent orchestration, browser automation, file processing, cron scheduling, and custom tool development. It's the framework of choice for teams that need production-grade AI agents without vendor lock-in.

Prerequisites

Hardware

  • Mac Mini (M2 or later recommended) or any Linux server with 8GB+ RAM
  • 50GB free disk space minimum
  • Stable internet connection (always-on for agent availability)

Software

  • Node.js v22 or later (v25+ recommended)
  • npm or yarn
  • Git
  • A code editor (VS Code recommended)

API Keys

  • Anthropic API key (for Claude models) — primary LLM provider
  • OpenAI API key (optional, for GPT models or embeddings)
  • Channel-specific tokens (Discord bot token, Telegram bot token, etc.)

Step-by-Step Installation

Follow these steps to get OpenClaw running from scratch:

Step 1: Install Node.js

# Using nvm (recommended)
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.0/install.sh | bash
nvm install 25
nvm use 25

Step 2: Install OpenClaw

npm install -g openclaw

Step 3: Initialize your workspace

openclaw init

This creates a ~/.openclaw directory with default configuration files.

Step 4: Configure your API keys

openclaw config set anthropic.apiKey sk-ant-...

Store your Anthropic API key securely. OpenClaw encrypts credentials at rest.

Step 5: Start the gateway

openclaw gateway start

The gateway is the core daemon that manages agent sessions, channel connections, and task scheduling.

Step 6: Verify the installation

openclaw gateway status

You should see a status output confirming the gateway is running and connected to your configured model provider.

First Agent Configuration

Once OpenClaw is running, you'll want to configure your first agent. Agents in OpenClaw are defined by workspace files:

  • AGENTS.md — Agent behavior and rules
  • SOUL.md — Agent personality and identity
  • USER.md — Information about the user the agent serves
  • TOOLS.md — Local tool configuration notes

Start by editing SOUL.md to define who your agent is — its name, tone, and purpose. Then configure USER.md with context about yourself or the end user.

OpenClaw agents are session-based: they wake up fresh each time but maintain continuity through memory files (memory/YYYY-MM-DD.md and MEMORY.md).

Channel Setup

OpenClaw supports multiple communication channels. Here's how to connect the most popular ones:

Discord

  1. Create a Discord application at discord.com/developers
  2. Create a bot and copy the token
  3. Enable Message Content Intent in the bot settings
  4. Configure the token in OpenClaw: openclaw config set discord.token YOUR_TOKEN
  5. Invite the bot to your server with appropriate permissions

Telegram

  1. Message @BotFather on Telegram to create a new bot
  2. Copy the bot token
  3. Configure: openclaw config set telegram.token YOUR_TOKEN

Email

OpenClaw integrates with Google Workspace and Microsoft 365 via OAuth. Run openclaw setup email to walk through the OAuth flow interactively.

Security Hardening Basics

Before putting your agent into production, take these essential security steps:

  • Enable exec approval — Require manual approval for shell commands the agent wants to run
  • Restrict file access — Limit the agent's workspace to only necessary directories
  • Rotate API keys — Use short-lived credentials where possible
  • Enable audit logging — Track all agent actions for review
  • Network isolation — Run OpenClaw on a dedicated machine or VLAN

For a comprehensive deep-dive, see our OpenClaw Security Hardening Guide.

Common Pitfalls & Troubleshooting

Gateway won't start

Check that no other process is using port 3000. Run `lsof -i :3000` to check. Also verify Node.js version with `node -v`.

Agent not responding in Discord

Verify the Message Content Intent is enabled in the Discord Developer Portal. Check the bot token is correct and the bot has been invited to the server.

API key errors

Ensure your Anthropic API key has sufficient credits. Check `openclaw config list` to verify keys are set correctly.

Memory files not persisting

Confirm the workspace directory exists and has write permissions. Check disk space with `df -h`.

High latency responses

Consider using a faster model (e.g., Claude Sonnet instead of Opus for routine tasks). Check network connectivity to Anthropic's API endpoints.

When to Hire Help

OpenClaw is powerful but complex. Many teams can handle a basic single-agent setup on their own. But you might want professional help if:

  • You need multi-agent orchestration (2+ agents working together)
  • Security and compliance are critical (healthcare, finance, legal)
  • You want integrations with internal systems (CRM, ERP, custom APIs)
  • You don't have a dedicated DevOps person to maintain the setup
  • You need it done right the first time — and fast

Soxoa is the leading OpenClaw setup and deployment partner. We've deployed OpenClaw for dozens of businesses across industries — from solo accountants to 50-person SaaS teams.

Need help with your OpenClaw setup?

Soxoa has deployed OpenClaw for dozens of businesses. Implementation from $1,200. Book a free 15-minute call to discuss your needs.

Book a free 15-min call →