Trae
AI-native integrated development environment (IDE) based on VS Code.
- VS Code-compatible AI-native IDE: A standalone editor that natively integrates AI features on top of the VS Code codebase. It is compatible with existing VS Code extensions while fully incorporating AI workflows. - Builder Mode: An agent-based development mode that automates the process from analyzing natural language requests to establishing step-by-step plans, modifying multiple files, providing live previews, and committing changes. It uses a planning-first approach to improve the accuracy of the initial attempt. - Chat Interface (General / Editor Inline): Supports both general question-and-answer mode and inline modification within the editor. It allows for multimodal input, such as screenshots and terminal output. - Built-in Multiple LLM Models: Supports direct switching and use of multiple models within the IDE, including Claude 3.5 Sonnet, Claude 3.7 Sonnet, GPT-4o, Doubao-1.5-pro, and DeepSeek. - MCP (Model Context Protocol) Support: Extends agent functionality to an MCP server, enabling features such as Figma design-to-code conversion, database integration, and external API integration. - trae-agent CLI (Open Source): An open-source agent that automates software engineering tasks based on natural language commands in a CLI environment, separate from the IDE. It supports Docker container isolation, trajectory logging, and YAML configuration. - Terminal Command Suggestions: Suggests terminal commands in the chat interface and allows for direct execution with a single click.
💻 System Requirements
Not required (AI inference is processed on a cloud server). When connecting to a local Ollama model using the trae-agent CLI, the VRAM requirements of that model apply.
IDE installation ~500MB, project and extension storage are separate.
⚡ Installation
### 4-1. Quick Start (IDE)
```
Download the OS-specific installation file from the official website (https://www.trae.ai/) and run it.
The installation process is the same as for VS Code, and it is compatible with existing VS Code extensions.
```
### 4-2. Installing trae-agent CLI
```bash
git clone https://github.com/bytedance/trae-agent.git
cd trae-agent
uv sync --all-extras
source .venv/bin/activate
# Create a configuration file
cp trae_config.yaml.example trae_config.yaml
# Set environment variables such as the API key, and then run
trae-cli run "Create a hello world Python script"
```
### 4-3. Main trae-agent CLI Commands
```bash
trae-cli run "Description of the natural language task" # Run a single task
trae-cli interactive # Interactive mode
trae-cli show-config # Check the current configuration
trae-cli run --docker-image python:3.12 "Task" # Run in Docker isolation
```🧬 Bio Use Cases
Natural Language-Based Full-Stack Prototyping
In Builder mode, generate HTML, CSS, JavaScript, and backend code in one go using natural language commands like "Create a customer feedback form with React + Express," and instantly check it with a live preview. Even non-developers can implement an MVP-level app within 30 minutes.
Multimodal UI Reproduction
Input screenshots or Figma design files through MCP to convert the UI into code. This reduces the designer-developer handoff time and minimizes the gap between design drafts and actual implementation.
CI/CD Automation Using trae-agent
Integrate the open-source CLI agent into the CI pipeline to automate code quality with natural language commands like "Write tests for this PR and fix lint errors." Docker-isolated execution ensures a secure sandbox environment.
📝 Update Notes
No update notes yet.
🧪 Related Code of Life
No related Code of Life posts yet.