Cursor IDE

What This Does

Cursor is an AI-native code editor. By pointing it at LLMAI's endpoint, you gain access to every model in the LLMAI catalog — GPT-5, Gemini, GLM — directly from Cursor's chat panel, inline edit, and code generation features. Token costs are drawn from your LLMAI balance the same as any other API call.

Setup

Step 1: Open the Settings File

1. Launch Cursor
2. Open the command palette: Cmd+Shift+P (macOS) or Ctrl+Shift+P (Windows / Linux)
3. Search for Open User Settings (JSON) and select it

Step 2: Add the LLMAI Configuration

Insert the following two fields into your settings.json:

{
  "cursor.general.openAIBaseURL": "https://api.llmai.dev/v1",
  "cursor.general.openAIAPIKey": "YOUR_LLMAI_API_KEY"
}

Retrieve your key from console.llmai.dev/keys.

Step 3: Configure via the UI (Alternative)

If you prefer the graphical settings panel:

1. Navigate to Cursor → Settings → Models
2. Find the API Key field and enter your LLMAI key
3. Find the Base URL or OpenAI API Base field and set it to https://api.llmai.dev/v1
4. Save; restart Cursor if prompted

Step 4: Pick Your Model

In Cursor's chat panel, type or select a model slug directly:

• gpt-5-4 — highest capability
• gpt-5-3-codex — optimised for code tasks
• gemini-3-1-pro — large context, multimodal
• glm-5-1 — balanced performance and reasoning
• glm-5-turbo — fast responses, low latency

Tip: Cursor rejects model names with dots. Use dash notation (e.g. gpt-5-4 instead of gpt-5.4). Both formats are equivalent — see Model Name Aliases for the full list.

Recommended Models for Coding

Troubleshooting

Cursor shows an authentication error

• Double-check that the API key was pasted without extra spaces or line breaks
• Verify the base URL is exactly https://api.llmai.dev/v1 (no trailing slash)

A model name is rejected

• Cursor does not accept dots in model names. Use the dash notation variant (e.g. gpt-5-4 instead of gpt-5.4)
• If a dot-notation slug is rejected, replace every . with -
• See the full alias table on the Models reference

Responses are slow

• Switch to a lower-latency model such as glm-5-turbo
• Check your remaining balance at console.llmai.dev/billing

Notes

• Cursor's UI layout may differ between versions. If you can't find a specific settings field, refer to Cursor's official documentation.
• All of Cursor's AI features — chat, inline edits, tab completions — function through LLMAI.
• Token usage is charged per interaction at the rate of whichever model is active.