Windsurf

What This Does

Windsurf is an AI-native code editor built by Codeium. It features deep AI integration including Cascade, an agent that can understand your entire codebase. By connecting Windsurf to LLMAI, you can use any model in the LLMAI catalog while managing costs through a single unified balance.

Setup

Step 1: Open Windsurf Settings

1. Launch Windsurf
2. Click the gear icon in the sidebar or press Cmd+, (macOS) / Ctrl+, (Windows/Linux)
3. Navigate to AI Providers or Cascade Settings

Step 2: Add a New Provider

1. Click Add Provider or + Add AI Provider
2. Select OpenAI Compatible from the provider list

Step 3: Configure LLMAI

Enter the following configuration:

Step 4: Save and Activate

1. Save the provider configuration
2. Select LLMAI as your active provider
3. Choose your preferred model from the model picker

Step 5: Alternative — Settings JSON

For advanced configuration, edit the settings file directly:

1. Open command palette: Cmd+Shift+P
2. Search for Open User Settings (JSON)
3. Add the configuration:

{
  "windsurf.aiProvider": "openai-compatible",
  "windsurf.openaiCompatibleBaseUrl": "https://api.llmai.dev/v1",
  "windsurf.openaiCompatibleApiKey": "YOUR_LLMAI_API_KEY",
  "windsurf.openaiCompatibleModel": "gpt-5.4"
}

Recommended Models for Windsurf

Troubleshooting

Provider not appearing in list

• Ensure you selected OpenAI Compatible, not plain OpenAI
• Save settings and restart Windsurf

Authentication errors

• Verify your API key is correct at console.llmai.dev/keys
• Re-paste the key without any extra spaces or newlines

Cascade doesn't work with LLMAI

• Cascade requires models with strong agentic capabilities — use gpt-5.4 for best results
• Some Cascade features may require specific model capabilities

Slow responses

• Use glm-5-turbo for faster responses
• Check your remaining balance at console.llmai.dev/billing

Notes

• Windsurf's UI evolves frequently — exact menu paths may differ between versions
• Cascade is Windsurf's flagship feature — it works best with capable models like GPT-5.4
• Token usage is logged to your LLMAI dashboard in real-time
• For tab completions, consider configuring a faster model separately if Windsurf supports it