TRAE

New

AI coding app with an IDE window and an Agent window, offering IDE and SOLO modes, custom agents with MCP, CUE code completion, and desktop, web and mobile clients that keep tasks in sync.

trae.ai

TRAE

Explore features, practical uses and pricing below.

TRAE is an AI development environment that pairs a conventional code editor with agents that plan and carry out multi-step work. The desktop app has two windows. The IDE window behaves like a familiar editor, with VS Code extensions, a terminal, debugging and Git. The Agent window is a task workspace for running several agent jobs at once, reviewing their changes and handing long tasks to the cloud. There are also web and mobile apps, and the official overview says projects, tasks and results stay in sync across all three.

IDE mode versus SOLO mode

Inside the IDE window you can switch between two ways of working. IDE mode keeps you in charge: you edit, run and debug while the AI helps on request. Developers coming from Cursor will recognise that loop. SOLO mode hands the lead to the AI. You describe the requirement in text or by voice, and TRAE plans the task, generates code, tests it, shows a preview and summarizes the changes. The docs frame the choice around task complexity and how much control you want. For a quick bug fix SOLO is usually more than you need, though it can save time on a throwaway prototype.

The parts that do the work

  • Agents search the codebase, write a multi-step plan and use tools to finish it. You can build custom agents with their own prompts, MCP servers and toolsets, for example a migration agent limited to database tools.
  • CUE is TRAE's completion engine. It handles inline completion, multi-line edits and next-edit prediction, and the docs list smart imports and reference renaming for Python, TypeScript and Go projects.
  • Context can include files, folders, snippets, terminal output, repositories, document sets and web pages.
  • Models include several built-in options, and you can add custom models with an API key.
  • AI code review summarizes and reviews uncommitted changes, single commits or branch diffs, with flowcharts and diff views.
  • Remote development works over Remote SSH or WSL.

Privacy mode and sandboxed commands

Two settings are worth turning on before you point TRAE at a work repository. With Privacy mode enabled, the docs say TRAE does not use your chats, code snippets or AI output for data analysis, product improvement or model training, and codebase files stay on your device. Sandboxed execution runs agent-generated commands in a restricted environment with file access controls and blocks on high-risk commands. Company rules on sending code to hosted models still apply.

Trying it on a real project

A reasonable first session looks like this. Open an existing TypeScript repo in the IDE window and make a small change by hand so you can judge CUE's suggestions on your own code. Then move to the Agent window and give an agent a bounded job, such as "add input validation to the signup form and write tests for the three failure cases," attaching the form component and its test file as context. When it finishes, run the AI review on the branch diff, run the tests yourself, and only then try SOLO mode on something disposable.

Plans and supported platforms

The pricing page lists a Free plan with limited usage, Auto mode only and 5,000 autocompletes a month. Pro is $20 a month, Pro+ $60 and Ultra $200. Paid plans express usage as a dollar amount of basic usage, add unlimited autocomplete and raise the number of concurrent cloud tasks, and yearly billing is advertised at a discount. Desktop downloads cover macOS 12 or later, Windows 10 and 11, and Linux through .deb and .rpm packages. For a lighter editor with fewer agent features, look at Zed.

Common questions from new TRAE users

Can I use my VS Code extensions?

The IDE window supports VS Code extensions, and TRAE also has its own extension store.

Does the free plan include every model?

No. The pricing page limits Free to Auto mode. Choosing from all models starts with Pro.

Do I have to install the desktop app?

No. TRAE Web runs tasks in a ready-made cloud environment in the browser, and the mobile app can assign and monitor tasks.

Similar tools

AWS-built agentic development environment that turns prompts into requirements, design and task specs, then runs agents in an IDE, CLI, web app or VS Code extension with hooks, steering files and MCP.
Fast collaborative code editor with AI agentic editing, inline assistance, edit prediction, and configurable model access.
AI Code EditorsSee website