Skip to main content

​
Extension Types

Qwen Code extensions support five main types of capabilities. Each type serves a different purpose and can be combined within a single extension.

​
MCP Servers

Model Context Protocol (MCP) servers add new tools that the AI can use to interact with external services, APIs, or perform specialized operations.

​
What are MCP Servers?

  • Standalone processes that communicate via stdio
  • Provide tools (functions) the AI can call
  • Can expose prompts and resources
  • Run continuously during your session

​
Use Cases

  • Fetch data from external APIs
  • Interact with databases
  • Control external services (deployment, monitoring)
  • Process files in specialized formats
  • Integrate with third-party tools

​
Example

{
  "name": "github-tools",
  "version": "1.0.0",
  "mcpServers": {
    "github": {
      "command": "node",
      "args": ["${extensionPath}${/}dist${/}server.js"],
      "cwd": "${extensionPath}",
      "env": {
        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
      }
    }
  }
}
Learn more: Creating MCP Server Extensions

​
Custom Commands

Custom commands are shortcuts for complex prompts that can be invoked with slash syntax.

​
What are Custom Commands?

  • Markdown files with optional YAML frontmatter
  • Can include template variables with {{args}}
  • Can execute shell commands with !{command}
  • Support nested directory structure for organization

​
Use Cases

  • Create project-specific workflows
  • Package common development patterns
  • Automate repetitive tasks
  • Build domain-specific helpers

​
Example

---
description: Search for a pattern in code and summarize findings
---

Please summarize the findings for the pattern `{{args}}`.

Search Results:
!{grep -r {{args}} .}
Invoked as: /fs:grep-code "TODO" Learn more: Creating Command Extensions

​
Skills

Skills are specialized capabilities that the AI can automatically invoke when relevant to the user’s request.

​
What are Skills?

  • Model-invoked capabilities (not user-invoked)
  • Defined in SKILL.md files with YAML frontmatter
  • Discovered automatically when extension is loaded
  • Available via /skills command to view/manage

​
Use Cases

  • Add domain-specific expertise
  • Provide specialized analysis capabilities
  • Enable automatic workflow handling
  • Extend AI reasoning for specific tasks

​
Example

---
name: code-analyzer
description: Analyzes code structure and provides insights about complexity and dependencies
---

# Code Analyzer

When analyzing code, focus on:
- Code complexity and maintainability
- Dependencies and coupling
- Potential performance issues
Learn more: Creating Skill Extensions

​
Subagents (Agents)

Subagents are specialized AI assistants configured for specific tasks with their own system prompts and tool access.

​
What are Subagents?

  • Markdown or YAML files defining agent behavior
  • YAML frontmatter specifies name, tools, model config
  • Content contains system prompt instructions
  • Appear in agent manager dialog

​
Use Cases

  • Create task-specific AI experts
  • Define consistent behavior patterns
  • Package reusable AI workflows
  • Share specialized assistants

​
Example

---
name: refactoring-expert
description: Specialized in code refactoring and improving code quality
color: blue
tools:
  - Read
  - Write
  - Grep
modelConfig:
  model: qwen3-coder-plus
---

You are a refactoring specialist focused on improving code quality.

Your expertise includes:
- Identifying code smells
- Applying SOLID principles
- Improving readability
Learn more: Creating Agent Extensions

​
Context Providers

Context providers supply persistent context to the model through a QWEN.md file that’s loaded in every session.

​
What are Context Providers?

  • Markdown file loaded into system context
  • Always available to the AI
  • Useful for instructions and documentation
  • Can describe extension capabilities

​
Use Cases

  • Provide usage instructions for extension tools
  • Define behavioral guidelines
  • Document project conventions
  • Add domain knowledge

​
Example

# Database Query Assistant

You have access to database query tools. When the user asks about data:

1. Use the `query_database` tool to fetch information
2. Format results in clear tables
3. Explain any complex queries

Always verify table names before querying.
Learn more: Creating Context Provider Extensions

​
Settings

Extension settings allow you to configure API keys, credentials, and other values securely.

​
What are Settings?

  • Defined in qwen-extension.json
  • Prompted during installation
  • Stored securely (sensitive values in keychain)
  • Passed as environment variables to MCP servers

​
Use Cases

  • Store API keys and tokens
  • Configure service endpoints
  • Set user preferences
  • Provide credentials

​
Example

{
  "name": "api-extension",
  "version": "1.0.0",
  "settings": [
    {
      "name": "API Key",
      "description": "Your API key for the service",
      "envVar": "MY_API_KEY",
      "sensitive": true
    },
    {
      "name": "API Endpoint",
      "description": "Base URL for the API",
      "envVar": "API_ENDPOINT",
      "sensitive": false
    }
  ]
}
Manage with: 
qwen extensions settings set my-extension "API Key"
qwen extensions settings list my-extension

​
Combining Extension Types

You can combine multiple types in a single extension: 
my-complete-extension/
├── qwen-extension.json       # Defines MCP servers and settings
├── QWEN.md                   # Provides context
├── commands/
│   ├── deploy.md            # Custom command
│   └── test.md              # Custom command
├── skills/
│   └── analyzer/
│       └── SKILL.md         # Skill
└── agents/
    └── expert.md            # Subagent

{
  "name": "my-complete-extension",
  "version": "1.0.0",
  "mcpServers": {
    "tools": {
      "command": "node",
      "args": ["${extensionPath}${/}dist${/}server.js"]
    }
  },
  "contextFileName": "QWEN.md",
  "commands": "commands",
  "skills": "skills",
  "agents": "agents",
  "settings": [
    {
      "name": "Service Token",
      "description": "Authentication token",
      "envVar": "SERVICE_TOKEN",
      "sensitive": true
    }
  ]
}
This creates a comprehensive extension with tools, commands, AI capabilities, and secure configuration.

​
Next Steps