Back to Tutorials
Clawdbot Tutorial10 minBeginner

Clawdbot Tutorial 01: Model Configuration and Switching

Getting errors when configuring models? The core issues are: command-line configuration and URL correction

Author: 收藏的AI工具箱|Published: 2026年2月1日|Original Link

Today I finally had time to configure the Macmini I bought specifically for Clawdbot, and刚好 MoltBook is hot to play with. Unexpectedly, I got stuck on model configuration for the entire afternoon.

Here's a note. If you're also troubled by this, you can refer to my approach.

Core Issues

The two core issues causing errors everywhere during model configuration: command-line configuration and URL correction

Use openclaw configure Command First

This is the easiest method that can solve most configuration problems.

Enter in terminal:

openclaw configure

It will ask you a few questions:

Local or Remote?

Choose Local

What to configure?

Choose Model

Enter your API Key after model selection

There's a Pitfall in Model Selection

Select the Corresponding Model

  • If you want to configure Minimax M2.1, select Minimax
  • If you want to configure Kimi K2.5, select moonshot AI

Note:

  • Kimi has a coding plan option, you can select it directly
  • Minimax doesn't have a coding plan option. Even if you bought the coding plan membership, select Minimax itself

After selecting the model, a bunch of options will pop up - Amazon, Google, etc. Don't panic, just press the arrow keys to scroll down, find the option that's already highlighted on the left (like Minimax or Kimi coding), and press Enter.

Domestic vs Overseas Versions - The Biggest Pitfall

⚠️ CRITICAL ATTENTION

Minimax has two options:

minimax-cn

Domestic version

minimax

Overseas version

If you bought the domestic coding plan membership, select cn. If overseas, select the one without cn.

Selected wrong? See below.

How to Manually Edit Configuration File

If openclaw configure still doesn't work, or you selected the wrong domestic/overseas version, you need to manually edit the configuration file.

Configuration file location:

/Users/你的用户名/.openclaw/openclaw.json

Open this file and find the baseURL line.

🔧 URL Correction

Domestic and overseas models use two different URLs. This is why most configuration failures happen.

For example, Minimax:

  • api.minimax.com - Domestic version
  • api.minimax.io - Overseas version

If it doesn't work after configuration, it's likely the URL is wrong. Go to the official documentation of the corresponding model to find the correct URL and change it.

Configure agents fallbacks

There's another parameter in openclaw.json to note: fallbacks in agents - model switching only takes effect here.

If you've done everything and still have problems, check if the model you switched to is written here. If not, you can add it.

Model Switching

After configuration, switching models is simple.

Enter in TUI interface:

/model

By the way, the TUI startup command is:

openclaw tui

Then search for the model name you want to switch to, like search "Kimi", the first result is Kimi for coding, press Enter to switch.

Recommendations for Switching Models

💡 Best Practice

Before switching models, it's recommended to use the /new command to open a new window first, then switch. This makes problems less likely.

The "no output" Issue - This Troubled Me the Longest

After switching models, you type "hello" and it returns no output. I thought the configuration failed, but it didn't.

What "no output" Means

"no output" means: the output is in another environment.

  • Web environment (claude.ai)
  • Telegram bot
  • Other environments you've configured

✅ Solution

So don't panic when you see no output, try other environments. If it works in other environments, it means the configuration is successful, just the output location is wrong.

Three Main Domestic Models Supported by openclaw

I configured all three, and they all work normally:

Kimi

Using domestic version

Minimax

Using overseas version

GLM

Using overseas version

Summary

1

Use openclaw configure command first

2

If that doesn't work, manually edit the baseURL in the configuration file

3

Switch models using /model command

4

Don't panic at no output, try other environments

Original link: 微信公众号