# Trae Setup Guide

> Connect LLMoxy to ByteDance's Trae IDE through a custom model, and do agent coding with Chat / Builder.

Website: [https://www.trae.cn (China)](https://www.trae.cn) · [https://www.trae.ai (International)](https://www.trae.ai)

Trae (pronounced /treɪ/) is ByteDance's AI IDE, built on VS Code, with Chat and Builder included. Besides the built-in models, you can use a custom configuration to send requests to LLMoxy.

Before you start, get an [LLMoxy token](https://en.llmoxy.com/docs/get-started/get-token). The **auto** group is recommended.

## Install

1. Open the official website and download the `.exe` or `.dmg` for your system (choose ARM for Apple Silicon).
2. On first launch, pick a theme and your language.
3. Optional: import extensions, keyboard shortcuts, and settings from VS Code or Cursor.
4. We recommend installing the `trae` command-line tool so you can open the current directory with `trae .`.

![Download Trae](https://en.llmoxy.com/docs-assets/migrated/tutorial-trae/download.png)

![Import from VS Code / Cursor](https://en.llmoxy.com/docs-assets/migrated/tutorial-trae/import.webp)

## Connect to LLMoxy

1. Open **Settings → Models**.
2. Click **Add Model**.
3. Choose **Custom Model** at the top (not a preset vendor such as DeepSeek or Alibaba Cloud).
4. Fill in the fields below and click Add. Trae first checks whether the key is valid.

![Open model management](https://en.llmoxy.com/docs-assets/migrated/tutorial-trae/models.png)

![Select custom model](https://en.llmoxy.com/docs-assets/migrated/tutorial-trae/add-model.webp)

**OpenAI-compatible (recommended; GPT / Gemini / most models):**

| Field | Value |
| --- | --- |
| API format | OpenAI Chat Completions |
| Custom request URL | `https://llmoxy.com/v1` |
| Full URL | **Off** |
| Model ID | For example `grok-4.7` |
| API key | LLMoxy token |

**Anthropic-compatible (Claude models):**

| Field | Value |
| --- | --- |
| API format | Anthropic Messages |
| Custom request URL | `https://llmoxy.com` |
| Full URL | **Off** |
| Model ID | For example `claude-opus-5-5`, `claude-fable-5-2` |
| API key | LLMoxy token |

> **💡 Tip**
> "Full URL" is off by default. When you turn it on, you must enter the complete path, for example `https://llmoxy.com/v1/chat/completions`. For GPT-5 / GPT-6, we recommend choosing the matching "model series" in the advanced settings so Trae adjusts the request fields automatically.

After it's added, turn off Auto at the bottom right of the chat input box and pick the model you just added from the list.

## How to use

The AI panel is on the right:

- **Chat**: explain code, find bugs, and ask questions with context via `@filename`.
- **Builder**: build a project from scratch in natural language. It creates files, installs dependencies, and runs a preview.

![Chat and Builder](https://en.llmoxy.com/docs-assets/migrated/tutorial-trae/assistant.png)

| Action | macOS | Windows |
| --- | --- | --- |
| Accept changes in the current file | `Command + Enter` | `Ctrl + Enter` |
| Reject changes in the current file | `Command + Backspace` | `Ctrl + Backspace` |
| Inline chat | `Command + I` | `Ctrl + I` |

VS Code extensions are also supported. If you prefer Cline, see the [Trae Cline Plugin Tutorial](https://en.llmoxy.com/docs/code-tutorials/trae-cline).

## FAQ

**The custom request URL stays red**
First make sure "Full URL" is off and the URL is `https://llmoxy.com/v1` (OpenAI) or `https://llmoxy.com` (Anthropic). Don't append `/chat/completions` to the base URL.

**Key check fails**
Check that the token has no spaces, the group is **auto**, and the model ID matches the console exactly.

**Builder reports a terminal error on Windows**
Go to Settings → Terminal profile and choose PowerShell 6 or later.

This is the markdown version of https://en.llmoxy.com/docs/code-tutorials/trae.
