Abstract

Trae is an AI-native integrated development environment (IDE) built by ByteDance, based on VS Code architecture, equipped with built-in AI chat assistance, automatic code completion, and end-to-end agent-driven autonomous development capabilities. It serves as a one-stop development tool supporting mainstream programming languages and full-cycle software project delivery. This guide systematically introduces the core SOLO mode of Trae IDE, covering workspace layout, activation steps, two core built-in agents (SOLO Coder for complex engineering development and SOLO Builder for rapid prototype creation), Plan task planning workflow, custom agent configuration, and cross-agent collaborative logic. Development teams managing multi-model LLM backend traffic can simplify unified request routing and credential management via an API gateway solution like Treerouter.

1. Basic Introduction to Trae IDE

Trae is a domestic AI development IDE derived from VS Code, designed for full-cycle engineering development. Its core value lies in agent-based autonomous workflow automation: developers only need to submit natural language requirements, and the internal AI agent will independently complete task decomposition, code writing, file modification, unit testing, and result preview without tedious manual multi-step operations.

1.1 Official Access & Installation

  1. Official website entry: https://www.trae.cn/
  2. Installation logic: The download button automatically matches the user’s operating system; users can also manually select installation packages for Windows, macOS, and other platforms.
  3. Supplementary learning resources: Official introductory tutorials and beginner documentation are available on the official site for environment setup and feature practice.

After finishing installation, launch the client, and the core mode switch button is located at the top left of the IDE window for quick entry to SOLO Mode.

2. Core Feature: SOLO Mode Overview

SOLO Mode is the flagship core capability of Trae IDE, with AI agents as the central executor to automate the entire development pipeline from requirement comprehension to deliverable preview.

2.1 Core Capabilities of SOLO Mode

  • Multi-dimensional requirement input: Natural language text dialogue, voice input, local file upload for context reference
  • Automatic task splitting: The built-in agent parses complex business demands into sequential executable subtasks
  • Full-link autonomous execution: Code generation, file editing, diff change inspection, build verification, and result preview run without human intervention
  • Multi-agent orchestration: Users can enable multiple custom agents to realize role-based collaborative development for large projects

2.2 How to Activate SOLO Mode

  1. Complete Trae IDE installation and open the workspace project folder
  2. Locate the mode toggle button at the top left of the IDE navigation bar labeled "Switch to SOLO Mode"
  3. Click the button to switch the entire workspace into the three-panel SOLO dedicated layout

2.3 Three-Panel SOLO Layout Specification

The SOLO workspace adopts a fixed three-column split layout, with clear functional division for each panel from left to right:

  1. Left panel: Task management dashboard — records all historical development tasks, displays task progress and execution logs, supports creating new blank tasks
  2. Middle panel: AI dialogue core panel — the main interaction area for communicating with SOLO agents, displaying agent thinking processes, generated code, and feedback results
  3. Right panel: Tool container panel — integrates built-in auxiliary tools including code editor, document reader, terminal, browser, Figma parsing, MCP service integration, and agent management configuration

3. SOLO Coder: Agent for Complex Software Engineering Projects

SOLO Coder is the dedicated built-in agent oriented toward medium-to-large complex software development, capable of end-to-end processing from demand analysis, architecture design, modular coding to iterative refactoring.

3.1 Core Functional Advantages

  1. Structured task planning: After receiving requirements, the agent generates a standardized execution plan for user confirmation before starting sequential implementation
  2. Multi-agent collaborative support: Users can configure multiple custom agents to divide specialized work, forming a virtual AI development team to accelerate project iteration
  3. Auto adaptive model scheduling: Defaults to Auto mode, which dynamically balances reasoning speed, hardware resource overhead, and output quality to match the optimal underlying LLM without manual parameter adjustment
  4. Full lifecycle code management: Supports global project refactoring, multi-file batch modification, and automatic code change difference inspection

3.2 Activation Steps for SOLO Coder

  1. Switch the IDE to SOLO Mode completely
  2. Locate the @ symbol at the bottom left of the middle dialogue input box
  3. Click the @ trigger to pop up the agent selection menu, then pick SOLO Coder to bind the current dialogue session

3.3 Custom Configuration for SOLO Coder

Users can edit available auxiliary tools, MCP server access permissions, and custom invoked agents for SOLO Coder:

  1. Hover the mouse over the SOLO Coder identifier on the right side of the dialogue panel to unfold the configuration icon
  2. Click the "Edit Tools" button to enter the dedicated configuration page
  3. Toggle switches to enable or disable built-in tools such as Search, file readers, and third-party extended capabilities

3.4 Custom Agent Invocation Logic

SOLO Coder supports calling self-defined independent agents to realize modular task decoupling:

  1. Built-in Search agent is enabled by default, for document retrieval and background data query to supplement reasoning context
  2. Multiple custom agents can be pre-configured with independent context memory; when handling multi-layer complex requirements, SOLO Coder automatically splits subtasks and invokes corresponding agents on demand
  3. Developers can explicitly specify target agents via @ mentions in dialogue prompts to force task assignment to designated specialized agents

4. Plan Mode: Controllable Task Planning Workflow

Plan Mode is an auxiliary sub-function under SOLO Coder, suitable for highly complex multi-stage projects that require human review of execution logic.

4.1 Activation Shortcuts

  • macOS: Option + P
  • Windows: Alt + P The Plan toggle switch is also placed at the top right of the dialogue panel for mouse click activation.

4.2 Execution Mechanism

  1. After receiving user requirements, SOLO Coder first generates a visualized structured task plan and displays it in the dialogue panel
  2. Users can manually edit the plan content to adjust subtask order, modify implementation logic, or delete redundant steps
  3. Click the "Execute" button to confirm the plan; the agent will carry out each subtask sequentially following the approved workflow
  4. If the generated plan deviates from business expectations, users can send natural language adjustment instructions to let the agent auto-revise the whole planning framework

5. SOLO Builder: Rapid Prototype Generation Agent

SOLO Builder is the second core built-in agent under SOLO Mode, focusing on ultra-fast end-to-end prototype development, especially for front-end Web application scaffolding without massive manual coding work.

5.1 Core Positioning

Different from SOLO Coder’s deep engineering iteration capability, SOLO Builder prioritizes speed and lightweight delivery: it can generate complete runnable Web prototypes from simple functional requirements in a short time, applicable to demand verification, demo development, and small internal tool construction.

5.2 Activation Operation

The activation flow is consistent with SOLO Coder: enter SOLO Mode, click the @ symbol in the dialogue input box, and select SOLO Builder from the pop-up agent list to bind the current conversation.

6. Core Collaborative Logic Between Agents

The two native agents (SOLO Coder, SOLO Builder) and user-defined custom agents support cross-call and context sharing within the SOLO workspace:

  1. Context isolation mechanism: Each agent maintains independent long-term memory, avoiding cross-contamination of irrelevant requirement information between different tasks
  2. Dynamic on-demand scheduling: The primary agent automatically judges subtask types and invokes matching auxiliary agents without extra user instruction
  3. Manual forced assignment: Users can use @[Agent Name] in prompts to lock the executor of specific subtasks, realizing precise manual control of task division

7. Conclusion

Trae IDE’s SOLO system reconstructs traditional software development workflows by shifting repetitive coding, file management, and task scheduling work to AI agents. SOLO Coder targets heavyweight formal engineering projects with controllable Plan planning and multi-agent collaboration, while SOLO Builder serves lightweight prototype and demo development scenarios with ultra-fast delivery speed.

The complete SOLO workflow covers requirement input, agent selection, task planning, autonomous execution, and tool-assisted verification, greatly lowering the threshold for full-stack development. For teams running mixed multi-LLM agent backends, centralized traffic management tools such as Treerouter can standardize model access and simplify cross-service operation maintenance. Developers can select matching agents and functional modes according to project scale, requirement complexity, and delivery speed demands to maximize the efficiency improvement brought by Trae’s native AI automation capabilities.