Skip to main content

​
Video walkthrough

Before you begin:
  • A running OpenClaw instance in Agent 37
  • Access to the instance terminal
  • A valid API key or token for the model provider you want to use

​
Why you might change the default model

Changing the default model is useful when you want to:
  • test a different provider
  • improve response quality
  • reduce cost or latency
  • switch to a model better suited to your workflow
OpenClaw lets you do this through the built-in configuration wizard, so you do not need to rebuild your full environment.

​
Step 1: Open your OpenClaw instance

  1. Sign in to the Agent 37 dashboard.
  2. Open your OpenClaw instances.
  3. Confirm your target instance is running.
Before you continue, check that the instance row shows a healthy running status.

​
Step 2: Open the terminal

Inside the instance row, click Terminal. This opens the command-line interface connected to your OpenClaw environment.

​
Step 3: Start the configuration wizard

Inside Terminal, run: 
openclaw configure
If an existing configuration is already present, OpenClaw will detect it and let you update the current setup. When prompted for the gateway location, the usual choice is: 
Local (this machine)

​
Step 4: Select the section to configure

The wizard will ask which parts of the configuration you want to update. You may see sections such as:
  • Workspace
  • Model
  • Web tools
  • Gateway
  • Daemon
  • Channels
To change the default AI model, select: 
Model

​
Step 5: Choose your model provider

Once the model section opens, OpenClaw will show the supported model providers. Depending on your setup, these may include:
  • OpenAI
  • Anthropic
  • MiniMax
  • Moonshot AI
  • Google
  • XAI
  • OpenRouter
  • Qwen
  • GLM
  • Copilot
  • Vercel AI Gateway
  • Venice AI
Choose the provider you want to use for the new default model.

​
Step 6: Choose the authentication method

Depending on the provider, OpenClaw may offer one or more authentication methods, such as:
  • API key
  • OAuth token
For example, Anthropic may offer both an API-key flow and a token-based flow. Most users will simply paste an API key.

​
Step 7: Paste your API key or token

After you choose the authentication method, OpenClaw will prompt you for the credential. At this step, you can:
  • paste the key or token
  • optionally give it a name
  • leave the name blank to use the default
When the credential is accepted, OpenClaw updates the configuration automatically. You may see a confirmation similar to: 
Updated ~/.openclaw/openclaw.json

​
Step 8: Select the default model

Once authentication is complete, OpenClaw will list the available models for that provider. Choose the model you want to use as the new default. From that point forward, your agent will use the selected model unless you change it again.

​
Step 9: Finalize the configuration

After the model selection, OpenClaw may continue through additional sections such as Workspace, Gateway, or Channels. You can update those sections if needed, or continue without making extra changes. When the wizard finishes, your new model configuration is active.

​
How to confirm the model change worked

To verify the new model is active:
  1. Open Chat or Web Chat
  2. Send a simple test prompt
  3. Confirm the agent responds normally
  4. Check terminal logs if you want to inspect the runtime behavior
If the agent responds correctly, the model change was successful.

​
Common mistakes

Watch for these common issues:
  • invalid API key or token
  • selecting a model not supported by the provider
  • exiting the wizard before the flow is complete
  • assuming the provider changed before model selection finished
If something goes wrong, rerun: 
openclaw configure

​
What to do next

Claude API

Configure Anthropic if you want to switch to Claude.

OpenAI API

Configure OpenAI if you want to switch to a GPT model.