Skip to content

trae

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

trae

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
README.md
README.md
 
 

README.md

Trae IDE

ByteDance's AI-first development tool with MCP and rules-based configuration.

Vendor: ByteDance | License: Proprietary | Runtime: Electron

Links

Installation

Download from https://www.trae.ai — available for Mac, Windows, and Linux (.deb/.rpm for x64).

Configuration Files

File Scope Purpose
.trae/rules/project_rules.md Project Project-level agent rules (see Rules section)
user_rules.md Global User-level agent rules — IDE-managed, auto-created via Settings UI (not manually placed)
.trae/mcp.json Project MCP server configuration
.trae/skills/{skill_name}/SKILL.md Project IDE-managed skill definitions (see Skills section)
.agents/skills/ Project Open agent skills standard — alternative skill location
.trae/commands/ Project Custom slash commands (v3.5.54+, up to 3 nesting levels)
IDE Settings UI Global All other settings

Note: .trae/rules/ is auto-created by the IDE if it does not exist. The IDE also auto-generates .trae/rules/git-commit-message.md for commit message rules.

Rules (.rules feature)

.rules is the feature name — not a file extension. Actual rule files use .md extension. Rules are Markdown text files that define persistent instructions for the agent, loaded during agent initialization and referenced during code completion and generation.

Project Rules

Stored at .trae/rules/project_rules.md (with the subdirectory .trae/rules/, not directly in .trae/). Multiple rules files are supported; the IDE creates and manages the .trae/rules/ folder.

Trae also reads .trae/rules/ folders in project subdirectories, enabling modular per-component rules.

As of v3.5.18 (January 2026), four application modes are available per rule:

Mode Behavior
Always Apply Applied in every agent session
Apply to Specific Files Applied when matching file globs are in context
Apply Intelligently Agent decides when to apply
Apply Manually Only applied when referenced with #Rule in chat 
# project_rules.md

## Code Style
- Use TypeScript strict mode
- Prefer functional components in React
- All async functions must have error handling

## Testing
- Write Jest tests for all new modules
- Maintain >80% coverage

Trae also reads AGENTS.md, CLAUDE.md, and CLAUDE.local.md from the project root as rule sources (toggleable in Settings > Rules).

User Rules

user_rules.md is the global personal rules file. It is IDE-managed and auto-created — do not place it manually. To create or edit it: open the AI dialog → Settings icon (upper right) → Rules → Personal Rules → click to create. The file applies across all projects.

Skills (v3.5.41+, March 2026)

Skills are on-demand reusable procedures — the key distinction from Rules is loading behavior:

Feature Rules Skills
Loading Always-on — loaded into every context On-demand — only loaded when explicitly invoked
Token cost Every session Only when used
Format .md files in .trae/rules/ SKILL.md files in skill directories
Invocation Automatic (per mode) Explicit invocation only

Skills are stored as SKILL.md files and can be placed in:

  • .trae/skills/{skill_name}/SKILL.md — IDE-managed location
  • .agents/skills/ — open agent skills standard (interoperable with other agents)

Skills allow teams to define reusable procedures (e.g., "run the test suite", "scaffold a new component") that only consume tokens when explicitly called, keeping routine contexts lean.

SOLO Mode

SOLO is Trae's autonomous multi-agent mode for end-to-end task execution. Unlike the standard agent (which is interactive and step-by-step), SOLO operates autonomously across the full development lifecycle:

  • Requirements analysis — breaks down tasks and plans implementation
  • Code generation — writes and edits files across the project
  • Terminal commands — runs builds, tests, and installs autonomously
  • Browser testing — interacts with a browser to verify functionality
  • Deployment — can execute deployment steps end-to-end

SOLO is designed for larger, self-contained tasks where you want minimal interruption. Full documentation: https://docs.trae.ai/ide/solo-mode

Slash Commands (v3.5.54+)

Custom slash commands can be defined in .trae/commands/. Up to 3 levels of nesting are supported, enabling structured command hierarchies (e.g., /project/setup/database).

Hooks

Note: Trae does not have a native pre/post tool use hook system. Lifecycle control is achieved through:

  1. MCP servers — expose custom tools and side effects
  2. Rules files — instruct the agent on what to do/avoid

MCP-Based Hook Patterns

Use MCP server tools to implement hook-like behavior:

// .trae/mcp.json
{
  "mcpServers": {
    "safety-guard": {
      "command": "node",
      "args": ["./mcp/safety-guard.js"]
    }
  }
}

The MCP server can implement a validate_command tool that the agent calls before running shell commands (by instruction in rules).

Built-in Agent Tools

Tool Description
Terminal Execute shell commands
File read/write Read and edit files
Web search Search the web
MCP tools From configured MCP servers

MCP Support (v1.3.0+)

Trae IDE v1.3.0 introduced MCP protocol support. Project-level MCP was added in January 2026.

MCP servers are configured in .trae/mcp.json at the project root. Global MCP servers can also be configured in IDE Settings.

Transport types supported: stdio, SSE, and Streamable HTTP (three transport types per official docs).

{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_TOKEN": "..." }
    }
  }
}

MCP connects external data sources and tools: databases, APIs, GitHub, Jira, etc.

Agent Configuration

Custom agents are configured via the IDE UI with:

  • Custom name and system prompt
  • Model selection
  • Available tools

Supported Models (2026)

Built-in models as of mid-2026:

Model Status
GPT-4o Available built-in
GPT-5 Available built-in
Gemini 2.5 Pro Available built-in
DeepSeek R1 Available built-in
DeepSeek V3 / Chat V3 Available built-in
Kimi-K2 Available built-in
Claude models Removed from built-in November 2025 (Anthropic regional restrictions); re-enable via custom provider with personal API key

See https://docs.trae.ai/ide/models for the current authoritative list.

Trae CN

See trae-cn/ — the Chinese-market version of Trae with the same architecture but different default models and regional endpoints.

Notes

  • Rules files use .md extension; .rules is the feature name, not a file extension.
  • Rules are loaded at agent initialization; updates may require restarting the agent session.
  • project_rules.md lives at .trae/rules/project_rules.md (note the rules/ subdirectory).
  • user_rules.md is IDE-managed (global); create via Settings UI, not by manually placing a file.
  • AGENTS.md, CLAUDE.md, and CLAUDE.local.md are read as rule sources (toggleable in Settings > Rules).
  • Skills (v3.5.41+) load on demand only — unlike Rules which are always-on — saving tokens on large contexts.
  • SOLO mode is the autonomous multi-agent mode for end-to-end task execution without step-by-step confirmation.
  • Custom slash commands live in .trae/commands/ (v3.5.54+, up to 3 nesting levels).
  • No native hook API; MCP is the primary extensibility path.
  • bytedance/trae-agent on GitHub is a separate CLI tool, not the Trae IDE. The official Trae IDE repo is Trae-AI/Trae.

Sources

Topic URL Fetched Label
Main docs https://docs.trae.ai 2026-06-13 [official]
Rules https://docs.trae.ai/ide/rules 2026-06-13 [official]
Skills https://docs.trae.ai/ide/skills 2026-06-13 [official]
SOLO mode https://docs.trae.ai/ide/solo-mode 2026-06-13 [official]
MCP protocol / transport types https://docs.trae.ai/ide/model-context-protocol 2026-06-13 [official]
Add MCP servers https://docs.trae.ai/ide/add-mcp-servers 2026-06-13 [official]
Models https://docs.trae.ai/ide/models 2026-06-13 [official]
Agent guide https://docs.trae.ai/ide/agent 2026-06-13 [official]
IDE settings overview https://docs.trae.ai/ide/ide-settings-overview 2026-06-13 [official]
Changelog https://docs.trae.ai/ide/changelog 2026-06-13 [official]
Trae IDE GitHub (official) https://github.com/Trae-AI/Trae 2026-06-13 [github]
trae-agent GitHub (separate CLI) https://github.com/bytedance/trae-agent 2026-06-13 [github]
v1.3.0 MCP + rules announcement https://traeide.com/news/6 2026-06-13 [third-party]