How to Install and Configure OpenClaw: A Self‑Hosted AI Assistant Gateway

This guide walks you through the overview, system requirements, step‑by‑step installation, custom model configuration, verification methods, core file locations, environment variables, hot‑reload options, common commands, and troubleshooting for OpenClaw, a self‑hosted AI assistant gateway.

Su San Talks Tech
Su San Talks Tech
Su San Talks Tech
How to Install and Configure OpenClaw: A Self‑Hosted AI Assistant Gateway

Overview

OpenClaw is a self‑hosted personal AI assistant gateway that provides a unified control plane for connecting chat applications (WhatsApp, Telegram, Discord, Slack, etc.) to AI programming agents while keeping all data on the local device.

System Requirements

Node.js ≥ 22 – verify with node --version Operating System: macOS or Linux (Windows is unstable; WSL2 is recommended)

AI provider API key – Anthropic (recommended), OpenAI, or a custom provider

Estimated installation time: 15 minutes

Installation

Homebrew (macOS only)

Skip this step on non‑macOS systems.

Check Homebrew: brew --version If not installed, run:

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

Install OpenClaw

The official script detects, installs, and initializes Node.js and runs the onboarding wizard:

curl -fsSL https://openclaw.ai/install.sh | bash

During the interactive setup you will be prompted to confirm actions (use the left arrow, select Yes, and press Enter). If the interface differs, abort with Ctrl+C and rerun the script.

After the script finishes you will be asked to create a gateway token, set permissions for ~/.openclaw (700), and create a session directory.

Installation step screenshots
Installation step screenshots

Custom Model Provider Configuration

Run the onboarding wizard to add a custom provider: openclaw onboard Select yes for personal use, choose QuickStart, then Use existing values. When prompted for a provider, choose Custom Provider and supply:

Base URL of the provider

API key generated by the provider

Endpoint compatibility: OpenAI‑compatible Model ID (e.g., gpt-5.3-codex)

Endpoint ID (auto‑generated, e.g., custom-claude-chiddns-com)

Skip optional channel and skill configuration, then restart the gateway.

Increase Context Window and Max Tokens

Default limits are often too low for large models. Adjust them with:

# Set context window
openclaw config set 'models.providers.custom-claude-chiddns-com.models[0].contextWindow' 400000
# Set max tokens
openclaw config set 'models.providers.custom-claude-chiddns-com.models[0].maxTokens' 128000
# Verify configuration
openclaw config get 'models.providers.custom-claude-chiddns-com.models[0]'

Verification

Check Service Status

# Gateway status
openclaw gateway status
# Full diagnostic
openclaw doctor
# Overall status
openclaw status

Access Control Interface

Terminal (recommended for servers): openclaw tui Desktop environment – launch the web dashboard: openclaw dashboard Or open http://127.0.0.1:18789/ in a browser.

Front‑end Debugging

openclaw gateway --port 18789
Dashboard screenshot
Dashboard screenshot

Core Files and Directories

~/.openclaw/openclaw.json

– main configuration file (JSON5 format) ~/.openclaw/workspace – agent workspace directory ~/.openclaw/.env – environment file storing API keys and other secrets

Environment Variables

OPENCLAW_HOME

– OpenClaw home directory OPENCLAW_STATE_DIR – state data storage directory OPENCLAW_CONFIG_PATH – path to the configuration file

Basic Configuration

The file ~/.openclaw/openclaw.json uses JSON5 and provides safe defaults. Manage it via the wizard:

# Launch configuration wizard
openclaw configure
# View current configuration
openclaw config get

Minimal Configuration Example

{
  "agents": {
    "defaults": {
      "workspace": "~/.openclaw/workspace"
    }
  }
}

Hot‑Reload Modes

Safe changes (e.g., channel settings) are applied instantly without restarting.

Infrastructure changes (e.g., port or authentication) require a gateway restart.

Common Commands Cheat Sheet

openclaw onboard

– run the onboarding wizard openclaw gateway – start the gateway openclaw gateway status – check gateway status openclaw dashboard – open the web console openclaw doctor – run diagnostic checks openclaw status – show overall system status openclaw logs --follow – tail logs in real time openclaw channels login – log in to chat channels openclaw configure – modify configuration openclaw config get – display current configuration openclaw agents add – add additional agent instances

FAQ

Q1: "EADDRINUSE" error on start?

Another gateway instance is using the port. Stop the other process or start on a different port, e.g.:

openclaw gateway --port 18790

Q2: Dashboard not reachable?

Confirm the gateway is running: openclaw gateway status If you see "device identity required", complete the device authentication flow.

When binding to a non‑loopback address, configure authentication first.

Q3: Features break after upgrade?

Upgrades may change defaults. Review the configuration and force reinstall of metadata:

openclaw config get
openclaw gateway install --force

Q4: How to update OpenClaw?

curl -fsSL https://openclaw.ai/install.sh | bash
CLIConfigurationNode.jsself‑hostedAI gatewayOpenClaw
Su San Talks Tech
Written by

Su San Talks Tech

Su San, former staff at several leading tech companies, is a top creator on Juejin and a premium creator on CSDN, and runs the free coding practice site www.susan.net.cn.

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.