Introduction to Extensions

Learn about Qwen Code extensions and how they expand the capabilities of your AI coding assistant

Extensions Overview

Qwen Code extensions package prompts, MCP servers, subagents, skills and custom commands into a familiar and user-friendly format. With extensions, you can expand the capabilities of Qwen Code and share those capabilities with others. They are designed to be easily installable and shareable.

What Can Extensions Do?

Extensions allow you to customize and extend Qwen Code in several ways:

• MCP Servers: Add new tools that the AI can use through the Model Context Protocol
• Custom Commands: Create shortcuts for complex prompts with /command syntax
• Skills: Add specialized capabilities that the AI can automatically invoke when relevant
• Subagents: Define specialized AI assistants for specific tasks
• Context Providers: Supply persistent context to the model for better responses
• Settings: Configure API keys and other credentials securely

Cross-Platform Compatibility

Qwen Code supports extensions from multiple ecosystems:

• Gemini CLI Extensions - Extensions from the Gemini CLI Extensions Gallery
• Claude Code Marketplace - Plugins from the Claude Code Marketplace
• Native Qwen Extensions - Built specifically for Qwen Code

Extensions and plugins from these platforms can be directly installed into Qwen Code. During installation, they are automatically converted to Qwen Code format:

• gemini-extension.json → qwen-extension.json
• claude-plugin.json → qwen-extension.json
• TOML command files → Markdown format
• Tool mappings and configurations are preserved

This cross-platform compatibility gives you access to a rich ecosystem without requiring extension authors to maintain separate versions.

Extension Structure

Every extension is a directory containing a qwen-extension.json manifest file:

my-extension/
├── qwen-extension.json    # Required manifest file
├── QWEN.md               # Optional context file
├── commands/             # Optional custom commands
│   └── my-command.md
├── skills/               # Optional skills
│   └── my-skill/
│       └── SKILL.md
├── agents/               # Optional subagents
│   └── my-agent.md
└── mcp-server/          # Optional MCP server code
    └── server.js

Installation Methods

You can install extensions from multiple sources:

From GitHub Repository

qwen extensions install https://github.com/owner/repo
# or
qwen extensions install owner/repo

From Claude Code Marketplace

qwen extensions install marketplace-name
# or
qwen extensions install marketplace-name:plugin-name

From Gemini CLI Extensions

qwen extensions install gemini-extension-repo-url

From Local Path

qwen extensions install /path/to/extension

For Development (Symlink)

qwen extensions link /path/to/extension

Runtime Management

Extensions can be managed at runtime within the interactive CLI:

Command	Description
/extensions	Manage all installed extensions
/extensions install <source>	Install an extension
/extensions explore [source]	Browse available extensions

CLI Management

You can also manage extensions using CLI commands:

qwen extensions install <source>     # Install extension
qwen extensions uninstall <name>     # Remove extension
qwen extensions enable <name>        # Enable extension
qwen extensions disable <name>       # Disable extension
qwen extensions update <name>        # Update extension
qwen extensions list                 # List all extensions
qwen extensions settings <name>      # Manage extension settings

Extension Lifecycle

On startup, Qwen Code:

1. Scans ~/.qwen/extensions/ for installed extensions
2. Loads each extension’s qwen-extension.json manifest
3. Registers MCP servers, commands, skills, and agents
4. Resolves conflicts with user/project commands
5. Makes everything available in the session

Changes made via runtime commands (slash commands) take effect immediately through hot-reloading. Changes made via CLI commands require restarting active sessions.

Enablement Scopes

Extensions can be enabled/disabled at two levels:

• User level (default): Applies across all workspaces
• Workspace level: Only applies to the current workspace

# Disable everywhere
qwen extensions disable extension-name

# Disable only in current workspace
qwen extensions disable extension-name --scope=workspace

Security Considerations

Extensions run with full access to your system:

• They can execute arbitrary code through MCP servers
• They can read and write files in your workspace
• They receive your API keys and credentials

Always review extension code before installing, especially from unknown sources. When installing, you’ll be prompted to acknowledge these security risks.

What’s Next?

Explore the different types of extensions you can create:

• Extension Types - Overview of all extension capabilities
• Creating Extensions - Step-by-step guide to building your first extension
• MCP Servers - Add custom tools via Model Context Protocol
• Custom Commands - Create reusable prompt shortcuts
• Skills - Build AI-invoked capabilities
• Subagents - Define specialized AI assistants