> ## Documentation Index
> Fetch the complete documentation index at: https://docs.leapx-hub.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Cursor

> Cursor Code Editor Integration Guide

## Introduction

Cursor is an AI-powered code editor. By integrating with **LeapX API**, you get powerful AI assistance for code generation, completion, explanation, and optimization. LeapX API now supports Cursor Agent capabilities (Auto Apply / multi-file editing, etc.) when configured properly.

## Quick Setup

<Warning>
  **Important**: Currently, Cursor has official restrictions for using third-party API keys (including LeapX API). You must meet the following requirements:

  * Must have an active **Pro Plan** (paid subscription). **Free Plan cannot use third-party API keys**
  * Requires VPN/proxy configuration to connect to third-party API services
  * This is an official Cursor policy restriction, not a limitation of LeapX API
</Warning>

### 1. Open Settings

Click the gear icon ⚙️ in the top right, select **Models**.

### 2. Configure API

* **OpenAI API Key**: Enter your LeapX API key
* **Override OpenAI Base URL**: Check and enter `https://api.leapx-hub.com/v1`
* Click **Verify** to validate

<Tip>
  - LeapX API is compatible with Cursor Agent (Auto Apply, etc.). To enable Agent, pick models that support Agent.
  - For GPT family, prefix model name with `openai/`. For example, use `openai/gpt-5.2` instead of `gpt-5.2`. Other providers keep their original model IDs.
</Tip>

### 3. Agent Mode Tips

* Use Agent-capable models (e.g., Claude, DeepSeek, Grok, and GPT with `openai/` prefix)
* Ensure Base URL is `https://api.leapx-hub.com/v1`
* On first enable, try a small change to verify auto-apply in Cursor sidebar
* If a model doesn’t support Agent, fall back to Chat and apply manually

## Recommended Models

| Type           | Model ID                   | Features                          |
| -------------- | -------------------------- | --------------------------------- |
| Coding         | `claude-sonnet-4-20250514` | Best coding ability               |
|                | `gpt-4.1`                  | Fast, strong overall              |
|                | `deepseek-v3`              | Great for Chinese, cost-effective |
| Cost-effective | `gpt-4.1-mini`             | Lightweight but capable           |
|                | `claude-3-haiku`           | Cheapest Claude                   |
|                | `gemini-2.5-flash`         | Google fast response              |
| Reasoning      | `o4-mini`                  | Top reasoning for coding          |
|                | `o3-mini`                  | Complex reasoning & algorithms    |

### Add Custom Models

Add in Cursor settings:

```
claude-sonnet-4-20250514
gpt-4.1
deepseek-v3
o4-mini
gemini-2.5-pro
```

## Usage Modes

### Chat Mode Workflow

1. **Generate code**: `Ctrl/Cmd + L` to open chat, describe requirements
2. **Apply code**: Copy and paste to target file, or use "Apply" button
3. **Iterate**: Continue conversation for modifications

### Alternatives Comparison

| Tool               | Agent Mode | Pros                         | Cons             |
| ------------------ | ---------- | ---------------------------- | ---------------- |
| Cursor             | ❌          | Elegant UI, great completion | No Agent mode    |
| Cline (VS Code)    | ✅          | Full Agent functionality     | Requires VS Code |
| RooCode (VS Code)  | ✅          | Multi-file editing           | Newer, evolving  |
| Continue (VS Code) | ✅          | Open source, customizable    | Complex setup    |

## Core Features

### Smart Code Completion

* **Tab completion**: Press Tab to accept suggestions
* **Multi-line**: Function-level code generation
* **Context-aware**: Precise suggestions based on project

### AI Chat

* `Ctrl/Cmd + K`: Open command palette
* `Ctrl/Cmd + L`: Open sidebar chat
* **Code explanation**: Select code and ask AI

### Code Editing

* **Generate**: Describe requirements, AI generates code
* **Refactor**: Get optimization suggestions
* **Fix errors**: AI helps locate and fix issues

## Shortcuts

| Shortcut       | Function                |
| -------------- | ----------------------- |
| `Ctrl/Cmd + K` | Open AI command palette |
| `Ctrl/Cmd + L` | Open AI chat sidebar    |
| `Tab`          | Accept suggestion       |
| `Esc`          | Cancel suggestion       |

## Tips

### Provide Clear Context

```javascript theme={null}
// @context: React component for user auth
// @requirements: Support OAuth2 login
// @constraints: Compatible with NextJS 13+
```

### Optimize Prompts

* **Bad**: `"Fix this function"`
* **Good**: `"Fix the floating point precision issue in calculateTotal function, ensure amounts are accurate to 2 decimal places"`

## Troubleshooting

<AccordionGroup>
  <Accordion title="Connection Timeout">
    * Check network connection
    * Verify API URL is `https://api.leapx-hub.com/v1`
    * Validate API key
  </Accordion>

  <Accordion title="Model Not Responding">
    * Check account balance
    * Try switching models
    * Restart Cursor
  </Accordion>

  <Accordion title="Poor Suggestions">
    * Provide more project context
    * Use more specific prompts
    * Switch to better performing model
  </Accordion>
</AccordionGroup>

## Best Practices

### Project Config

Create `.cursor-settings.json` in project root:

```json theme={null}
{
  "model": "gpt-4.1",
  "temperature": 0.7,
  "contextFiles": ["README.md", "package.json"],
  "rules": [
    "Use TypeScript strict mode",
    "Follow ESLint rules",
    "Add appropriate comments"
  ]
}
```

### Code Review

```
Please review this code, focus on:
1. Performance issues
2. Security vulnerabilities
3. Code standards
4. Best practices
```

## About Agent Mode

For auto multi-file editing and complex refactoring, consider:

* **Cline**: VS Code plugin with full Agent functionality
* **RooCode**: Emerging VS Code AI Agent plugin