OpenClaw integration guide

Open sourceUpdated: 2026-03-10

Introduction

OpenClaw (formerly Clawdbot / Moltbot) is an open-source personal AI assistant platform that lets you talk to AI through several messaging channels. With SeaWhale AI you can quickly connect Claude, Gemini, Qwen and other models:

Multiple interfaces: command line (TUI), web UI, DingTalk and more
Wide model choice: works with Claude, Gemini, Qwen, DeepSeek and other leading models
OpenAI-compatible API: connects through the standard SeaWhale AI API format
Flexible cost control: pay as you go, with free credit for new accounts

Important

OpenClaw was previously named Moltbot / Clawdbot, and some commands have not been fully renamed yet. If a new command such as openclaw dashboard fails with command not found: openclaw, fall back to the older equivalent (clawdbot dashboard or moltbot dashboard).

Supported models

Through the SeaWhale AI OpenAI-compatible API, OpenClaw supports:

Category	Models
Claude family	Claude Sonnet 4.5, Claude Opus 4.5, Claude Haiku 4
Gemini family	Gemini 2.5 Pro, Gemini 2.5 Flash
Qwen family	Qwen Max, Qwen Plus, Qwen Flash, Qwen Coder, Qwen3
Others	DeepSeek, Kimi, GLM, MiniMax

For the full model list and pricing, see the model list.

Recommended plan

Plan	Description
Pay as you go	Pay for what you use, with no commitment

Before you begin

Before configuring anything, make sure you have:

A SeaWhale AI account (if you do not have one, sign up in the SeaWhale AI console)
An API key generated on the API management page
Enough balance or free credit on your account

Installing OpenClaw

Requirements

OpenClaw requires Node.js 22 or later. Check your version with:

bash

node --version

If Node.js is missing or too old, download it from the Node.js website.

Installation methods

macOS/Linux (recommended)Windowsnpm global install

bash

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

powershell

iwr -useb https://openclaw.ai/install.ps1 | iex

bash

npm install -g openclaw@latest

Onboarding wizard

OpenClaw starts its configuration wizard automatically after installation. You can also launch it manually:

bash

openclaw onboard

Recommended answers for the initial setup:

Prompt	Recommended answer
I understand this is personal-by-default and shared/multi-user use requires lock-down. Continue?	Choose Yes
Onboarding mode	Choose QuickStart
Model/auth provider	Choose Skip for now (configure SeaWhale AI later)
Filter models by provider	Choose All providers
Default model	Choose Keep current
Select channel (QuickStart)	Choose Skip for now (configure channels later)
Search provider	Choose Skip for now
Configure skills now?	Choose No
Enable hooks?	Press space to select, then Enter to continue
How do you want to hatch your bot?	Choose Do this later

Configuring SeaWhale AI models

After initial setup, configure a SeaWhale AI model to enable conversations.

Configuration parameters

Parameter	Value
Base URL	`https://api.atalk-ai.com/v2`
API Key	Your SeaWhale AI credential, used for authentication and billing
Model ID	The model to use, e.g. `claude-sonnet-4-6` or `qwen3.5-plus`

Option 1: web console (recommended)

1. Start the web console

bash

openclaw dashboard

Your browser opens the console automatically (usually at http://127.0.0.1:18789). If it does not, open that address manually.

openclaw dashboard

2. Open the configuration page

In the left sidebar, go to Config > RAW.

openclaw configuration

3. Add the SeaWhale AI configuration

Copy the configuration below, replacing the existing "agents": {...} section, and substitute your SeaWhale AI API key for YOUR_API_KEY:

Important

Replace YOUR_API_KEY with your SeaWhale AI API key.
The example uses the claude-sonnet-4-6 and qwen3.5-plus models — substitute others as needed.
To use a different model, add its definition under providers.haijing.models (with reasoning set to false) and add a "haijing/MODEL_ID": {} entry under agents.defaults.models.

json


"models": {
  "mode": "merge",
  "providers": {
    "haijing": {
      "baseUrl": "https://api.atalk-ai.com/v2",
      "apiKey": "YOUR_API_KEY",
      "api": "openai-completions",
      "models": [
        {
          "id": "claude-sonnet-4-6",
          "name": "claude-sonnet-4-6",
          "reasoning": false,
          "input": ["text", "image"],
          "contextWindow": 200000,
          "maxTokens": 65536
        },
        {
          "id": "qwen3.5-plus",
          "name": "qwen3.5-plus",
          "reasoning": false,
          "input": ["text", "image"],
          "contextWindow": 1000000,
          "maxTokens": 65536
        }
      ]
    }
  }
},
"agents": {
  "defaults": {
    "model": {
      "primary": "haijing/claude-sonnet-4-6"
    },
    "models": {
      "claude-sonnet-4-6": {},
      "qwen3.5-plus": {}
    }
  }
}

4. Save and apply

Click Save in the top right, then click Update to apply the configuration.

Option 2: edit the config file

You can also edit openclaw.json directly, pasting in the JSON above and saving.

Starting a conversation

Web interface

Run this in your terminal to chat through the web UI:

bash

openclaw dashboard

openclaw configuration

Command line (TUI)

Run this to chat from the terminal:

bash

openclaw tui

openclaw configuration

Switching models

Temporarily (current session only)

Type this command in the TUI:

/model qwen3.5-plus

The response model set to qwen3.5-plus confirms the switch.

Permanently (all new sessions)

Change agents.defaults.model.primary in your config file to the target model:

json

{
  "agents": {
    "defaults": {
      "model": {
        "primary": "haijing/claude-sonnet-4-6"
      }
    }
  }
}

Troubleshooting

Q1: Model not found, or empty replies

Check each of the following:

The model ID is spelled correctly
The provider name in the config matches the reference (if the config uses haijing, the reference must be haijing/claude-sonnet-4-6)
reasoning must be set to false
A 400 status code with no response body means your balance is exhausted — top up and try again

openclaw configuration

Q2: "command not found" when running `openclaw`

Possible causes and fixes:

Cause	Fix
Installation failed	Run `openclaw --version`; a version number means it installed
Command not yet renamed	Try the older `moltbot` or `clawdbot` command instead
Version too old	Reinstall the latest: `npm install -g openclaw@latest`

OpenClaw integration guide ​

Introduction ​

Supported models ​

Recommended plan ​

Before you begin ​

Installing OpenClaw ​

Requirements ​

Installation methods ​

Onboarding wizard ​

Configuring SeaWhale AI models ​

Configuration parameters ​

Option 1: web console (recommended) ​

Option 2: edit the config file ​

Starting a conversation ​

Web interface ​

Command line (TUI) ​

Switching models ​

Temporarily (current session only) ​

Permanently (all new sessions) ​

Troubleshooting ​

Q1: Model not found, or empty replies ​

Q2: "command not found" when running openclaw ​

Related resources ​

Recommended reading ​

OpenClaw integration guide

Introduction

Supported models

Recommended plan

Before you begin

Installing OpenClaw

Requirements

Installation methods

Onboarding wizard

Configuring SeaWhale AI models

Configuration parameters

Option 1: web console (recommended)

Option 2: edit the config file

Starting a conversation

Web interface

Command line (TUI)

Switching models

Temporarily (current session only)

Permanently (all new sessions)

Troubleshooting

Q1: Model not found, or empty replies

Q2: "command not found" when running `openclaw`

Related resources

Recommended reading