Open main menu

How to Change Default AI Model in OpenClaw

This guide explains how to change the default AI model in OpenClaw. Learn how to configure a new model provider, add authentication keys, and set a new default model for your AI agent.

Table of Contents
Do not index
 
AI agents are only as powerful as the model running behind them. Sometimes you may want to switch models — maybe to test a different provider, improve response quality, or optimize costs.
OpenClaw makes this relatively simple through its configuration wizard inside the terminal. Instead of rebuilding your entire setup, you can update the model provider and set a new default model in just a few steps.
In this guide, you’ll learn how to change the default AI model in OpenClaw, using the built-in configuration process available in your Agent37 instance.
If you’re using an OpenClaw instance, you can access it directly through your dashboard:
👉 https://www.agent37.com/dashboard

Watch the Video Tutorial

If you prefer to follow along visually, you can watch the walkthrough here: 👉 https://www.loom.com/embed/4a7101cc1b854c00a2dbb89760bec156
The steps below explain the exact process shown in the video.

Step 1: Open Your OpenClaw Instance

First, log into your Agent37 workspace.
From the dashboard, navigate to your OpenClaw instances section where you’ll see your running environments.
You should see information like:
  • Instance ID
  • Tier (Basic, etc.)
  • Status (Running)
  • Free credits remaining
  • Actions like Chat, Terminal, and Restart
Before continuing, make sure your instance status shows:
Running
If the instance is stopped, start or restart it before proceeding.

Step 2: Open the Terminal

To modify configuration settings, you’ll need access to the OpenClaw terminal.
Inside your instance row, click:
Terminal
This launches the command-line interface connected to your OpenClaw environment.
You’ll see the OpenClaw configuration screen appear in the terminal.

Step 3: Start the Configuration Wizard

Inside the terminal, OpenClaw detects your current configuration.
You’ll typically see something like:
OpenClaw configure
Existing config detected
gateway.mode: local
This means your environment already has a configuration, and you’re about to modify it.
Next, OpenClaw asks:
Where will the Gateway run?
Usually the default option appears:
Local (this machine)
For most users, leaving this unchanged is perfectly fine.

Step 4: Select Sections to Configure

The wizard will ask which configuration sections you want to update.
You may see options such as:
  • Workspace
  • Model
  • Web tools
  • Gateway
  • Daemon
  • Channels
To change the AI model, select:
Model
This tells OpenClaw that you want to update the model provider configuration.

Step 5: Choose Your Model Provider

Once the model section opens, you’ll see the Model/Auth Provider list.
Some available providers include:
  • OpenAI
  • Anthropic
  • MiniMax
  • Moonshot AI (Kimi K2.5)
  • Google
  • XAI (Grok)
  • OpenRouter
  • Qwen
  • GLM
  • Copilot
  • Vercel AI Gateway
  • Venice AI
You can choose any provider supported by your setup.
In the walkthrough example, Anthropic is selected.
After selecting the provider, OpenClaw will ask how you want to authenticate.

Step 6: Choose Authentication Method

Depending on the provider, OpenClaw may offer two authentication options:
  • API key
  • OAuth token
For Anthropic, you may see:
Anthropic auth method
• Anthropic token (paste setup-token)
• Anthropic API key
You can choose whichever method you prefer.
Most users simply paste their API key.

Step 7: Paste Your API Key or Token

Once authentication is selected, OpenClaw prompts you to paste your key.
Example:
Paste Anthropic setup-token
Token name (blank = default)
You can:
  • Paste the token
  • Optionally give it a name
  • Or leave the name blank to use the default
After confirming the key, OpenClaw updates the configuration file automatically.
You’ll see something like:
Updated ~/.openclaw/openclaw.json
This confirms the new provider settings were saved.

Step 8: Select the Default Model

After authentication, OpenClaw lists available models from the provider.
For example:
Anthropic OAuth models
anthropic/claude-haiku-4-5
anthropic/claude-opus-4-6
From this list, choose the model you want to set as the default AI model.
Once selected, OpenClaw stores the model preference in the configuration.
From this point forward, your agent will use the new model by default.

Step 9: Finalize the Configuration

After completing the model setup, OpenClaw may display additional configuration sections like:
  • Workspace
  • Web tools
  • Gateway
  • Channels
You can either configure them or continue without changes.
When the wizard finishes, the configuration is saved automatically.
If you see output related to the Control UI, such as:
Web UI: http://127.0.0.1:18789/
Gateway WS: ws://127.0.0.1:18789
This simply indicates that the control interface is available.
Your new model configuration is already active.

How to Confirm the Model Change Worked

To verify that your new AI model is active:
  1. Open the Chat interface
  1. Send a simple prompt
  1. Observe the response behavior
  1. Check the terminal logs if needed
If the agent responds correctly, the model change was successful.

Common Mistakes When Changing the Default Model

A few small mistakes can prevent the model from working correctly.
Some common issues include:
  • Entering an invalid API key
• Choosing a model not supported by the provider
• Not completing the configuration wizard
• Forgetting to save the configuration
If something goes wrong, simply rerun the configuration process.

Final Thoughts

Changing the default AI model in OpenClaw is straightforward once you understand how the configuration wizard works.
The process typically involves:
  1. Opening the instance terminal
  1. Running the configuration wizard
  1. Selecting the model provider
  1. Adding authentication credentials
  1. Choosing the model you want to run
Once configured, your OpenClaw agent will immediately begin using the new model.
This flexibility is one of the strengths of OpenClaw — it allows you to experiment with different AI providers and choose the one that best fits your workflow.
If you’re running OpenClaw through Agent37, you can manage everything directly from the dashboard:
👉 https://www.agent37.com/dashboard
From there, you can launch instances, configure providers, and continuously optimize your AI agent setup.