Cursor

Cursor

Configure Cursor, the AI-powered code editor, to use agentgateway as its LLM backend.

Overview

Cursor supports custom OpenAI-compatible endpoints through its model configuration settings. This allows you to route all Cursor AI requests through agentgateway.

Prerequisites

• agentgateway running and accessible (e.g., http://localhost:3000)
• A configured LLM backend in agentgateway
• Cursor installed (version 0.30+)

Configuration

Method 1: Settings UI

1. Open Cursor Settings:

   • macOS: Cmd + , or Cursor → Settings
   • Windows/Linux: Ctrl + , or File → Preferences → Settings

2. Navigate to Models section

3. Click Add Model or Configure Custom Model

4. Enter your agentgateway details:

   • API Base URL: http://localhost:3000/v1 (or your gateway URL with /v1 path)
   • API Key: Your gateway API key, or anything if no authentication
   • Model Name: The model configured in your agentgateway backend (e.g., gpt-4o-mini, claude-sonnet-4-20250514)

5. Save and restart Cursor

Method 2: Settings JSON

Cursor stores configuration in a JSON file. You can edit this directly for more control.

1. Open Command Palette:

   • macOS: Cmd + Shift + P
   • Windows/Linux: Ctrl + Shift + P

2. Type Preferences: Open User Settings (JSON) and select it

3. Add your custom model configuration:

{
  "cursor.models": [
    {
      "name": "agent-gateway",
      "apiBase": "http://localhost:3000/v1",
      "apiKey": "anything",
      "model": "gpt-4o-mini"
    }
  ]
}

4. Save the file and restart Cursor

Multi-provider setup

You can configure multiple gateway backends as different models in Cursor:

{
  "cursor.models": [
    {
      "name": "gateway-openai",
      "apiBase": "http://localhost:3000/v1",
      "apiKey": "anything",
      "model": "gpt-4o-mini"
    },
    {
      "name": "gateway-anthropic",
      "apiBase": "http://localhost:3001/v1",
      "apiKey": "anything",
      "model": "claude-sonnet-4-20250514"
    }
  ]
}

Then switch between them in Cursor’s model selector.

Using environment variables

Cursor respects OpenAI environment variables. Set these before launching Cursor:

export OPENAI_API_BASE=http://localhost:3000/v1
export OPENAI_API_KEY=anything
cursor .

Example agentgateway configuration

Here’s a complete gateway configuration for Cursor with OpenAI backend:

# yaml-language-server: $schema=https://agentgateway.dev/schema/config
binds:
- port: 3000
  listeners:
  - routes:
    - policies:
        backendAuth:
          key: $OPENAI_API_KEY  # Your OpenAI API key
      backends:
      - ai:
          name: openai
          provider:
            openAI:
              model: gpt-4o-mini

Verification

Test your configuration:

1. Open a file in Cursor
2. Open the Cursor chat panel (Cmd + L on macOS, Ctrl + L on Windows/Linux)
3. Ask a question: “What is this file about?”
4. Cursor should respond using your agentgateway backend

Troubleshooting

“Failed to fetch” error

• Verify agentgateway is running: curl http://localhost:3000/v1/models
• Check the base URL includes /v1 path
• Ensure no firewall blocking the connection

“Invalid API key” error

• If agentgateway has no auth configured, set API key to any placeholder value (anything)
• If using backendAuth policy, ensure your gateway has valid provider credentials

Model not found

• Verify the model name matches your agentgateway backend configuration
• Check agentgateway logs for backend connection errors

Related documentation

• agentgateway LLM Configuration
• Backend Authentication