What is TRAE

TRAE is an AI tool brand from ByteDance, with the tagline “The Real AI Engineer”. It currently has two product lines:

  • TraeCode: an AI coding tool for developers, available as an IDE, a plugin and a CLI. TRAE calls it “your AI development engineer”.
  • TraeWork: an AI platform for everyday office work, on web, desktop and mobile, with Work, Code and Design modes for building slide decks, analysing data, doing deep research and writing documents.

This article focuses on TraeCode. The official docs describe it as a development tool with AI built in, covering the whole workflow from writing code and understanding a project to debugging, running and managing changes. Its defining idea is giving you both kinds of control: drive every step yourself as in a traditional IDE, or hand a complex task to an AI agent to plan and carry out.

The Chinese site is trae.cn and the international site is trae.ai.

AI coding tools tend to follow one of two approaches. One adds an AI assistant to a traditional editor and keeps the human in charge. The other lets AI lead development while the human sets requirements and signs off. TraeCode puts both in one product, as IDE mode and SOLO mode, and lets you switch based on the task. Fixing a production bug, you may want control over every line. Prototyping a new feature quickly, SOLO mode often gets there with less effort.

Key features

Two development modes. TraeCode has an IDE mode and a SOLO mode, and you can switch depending on the task:

  • IDE mode keeps the familiar editor, terminal, debugger, extensions and source control, for work where you want fine control over each change.
  • SOLO mode puts AI in the lead. Describe what you want in text or by voice, and the AI plans the work and moves through code generation, testing, preview and a summary of the changes.

Agents. Define a task in plain language and an agent understands it, searches the codebase, makes a multi-step plan and calls tools to get it done. You can also create custom agents with their own prompts, MCP servers and tool sets for particular kinds of work.

CUE completion. CUE handles code completion, chained completion, multi-line edits and predicting and jumping to the next edit. In Python, TypeScript and Go projects it can also help with imports and renaming references.

Rich context. Files, folders, snippets, terminal output, repositories, documentation sets and web pages can all be given to the AI as context.

Multiple models. Several models are built in, and you can bring your own through an API key.

A full toolchain. Editing, running and debugging for mainstream languages and frameworks; built-in Git with AI-written commit messages; AI code review that summarises uncommitted changes, single commits or branch diffs with summaries, flowcharts and diff views; an extension ecosystem; and remote development over Remote SSH or WSL.

Security controls. With privacy mode on, TraeCode does not use your conversations, code snippets or AI output for analytics, product improvement or model training, and your codebase stays on your device. Sandboxed execution runs agent-generated commands in a restricted environment, with file access controls and blocking of high-risk commands.

How to use it

  1. Install. Go to trae.ai (or trae.cn), download TraeCode and sign in.
  2. Open a project. Open an existing repository or start from an empty folder.
  3. Pick a mode. Stay in IDE mode if you want to write the code yourself with CUE and side-panel chat. Switch to SOLO mode to hand over a whole requirement in plain language.
  4. Add context. Reference relevant files, terminal errors or documentation links so the agent understands the task.
  5. Review and commit. Use AI code review to read the summary and diff, then commit with an AI-generated message.

For sensitive code, check your privacy mode and sandbox settings before you start.

Use cases

  • Starting a project from scratch: describe it and let an agent break it down, generate code, run checks and show a preview.
  • Maintaining an existing codebase: find the relevant files, understand dependencies, refactor or debug.
  • Everyday productivity: let CUE handle completions, multi-line edits and renames.
  • Team-specific workflows: build custom agents for fixed processes such as code review or test generation.
  • Office work: use TraeWork from the same brand for documents, slides and data analysis.

FAQ

What is the difference between TraeCode and TraeWork? TraeCode is for software development. TraeWork is an AI platform for office tasks such as documents, slides, data analysis and research. They are separate downloads.

Should I use IDE mode or SOLO mode? Use IDE mode when you want fine control, or when making targeted changes in a large existing project. Use SOLO mode when the requirement is well defined and you want AI to drive. You can switch at any time.

Will my code be used to train models? With privacy mode on, TRAE states that conversations, code snippets and AI output are not used for analytics, product improvement or model training, and codebase files stay on your device.

How is it priced? TRAE uses a credit-based billing model with several membership tiers, plus a “fast pass” option to speed up requests at peak times. See the official pricing page for current prices.

Official website