Qwen Orchestrator — Multi-Agent AI Development Team for Qwen Code

English · Español · 中文 · 日本語 · 한국어 · Português · Français · العربية

🤖 The enterprise-grade multi-agent AI orchestration extension exclusively for Qwen Code CLI

Turn your AI coding assistant into a full software development department — 24 specialized agents, 82 professional skills, 6 slash commands, persistent memory, and MCP tool integration.

Author: Omar-Obando · License: MIT · Version: 0.0.6

---

📑 Table of Contents

• What is Qwen Orchestrator?
• Quick Install
• Quick Start
• Key Features
• Agent Team (24 Specialized Agents)
• Skills (82 Professional Skills)
• Commands (6 Slash Commands)
• How It Works
  • Execution Flow
  • User Clarity: Ask Before Building
  • Architecture Diagram
• Installation Guide
  • Quick Install (Recommended)
  • Manual Install
  • MCP Tools Server (Optional)
  • Troubleshooting
• Model Configuration
• CLI Mode Recommendations
• Advanced Tools
  • Inter-Agent Communication
  • Real-Time Monitoring
  • Scheduled Tasks
• Integrations
  • Context7 (Optional)
  • MCP Memory Server
  • UI/UX Skills (Recommended)
• Quality Standards
• Project Structure
• First Usage Examples
• Documentation
• Author
• Keywords

---

⚡ What is Qwen Orchestrator?

Qwen Orchestrator is a powerful extension built exclusively for Qwen Code — the open-source AI coding CLI by Alibaba. It transforms a single AI coding assistant into a coordinated team of 24 specialized AI agents that work together like a professional software development department.

Why Qwen Orchestrator?

Imagine having an entire engineering team at your fingertips: a Commander that orchestrates, a Planner that architects, Frontend and Backend Developers that code in parallel, a Reviewer that gatekeeps quality, a QA Engineer that tests, a Cybersecurity Engineer that audits vulnerabilities, a DevOps Engineer that deploys — and 16 more specialists, all coordinated automatically.

No other AI coding extension provides this level of professional multi-agent orchestration.

Built Exclusively For

Qwen Code by QwenLM / Alibaba — The open-source AI-powered coding assistant CLI that supports multiple LLM providers (Qwen, DeepSeek, OpenAI, Anthropic, local models). Qwen Orchestrator is a community-built extension and is not affiliated with or endorsed by Alibaba.

⚠️ This extension ONLY works with Qwen Code CLI. It is NOT a standalone tool, NOT a VS Code extension, and NOT compatible with other AI coding assistants. If you don't have Qwen Code installed, install it first.

⚠️ IDE Compatibility Notice: This extension has NOT been tested with Visual Studio, VS Code extensions, JetBrains IDEs, or any integrations outside of the Qwen Code CLI. It is designed exclusively for the Qwen Code terminal/CLI experience.

---

🚀 Quick Install

qwen extensions install https://github.com/Omar-Obando/qwen-orchestrator

That's it! The extension automatically registers all 24 agents, 82 skills, 6 commands, and the MCP Memory Server.

📖 See the full Installation Guide for manual install, MCP tools server setup, and troubleshooting.

---

🎯 Quick Start

Once installed, just type in your Qwen Code CLI:

/orchestrator Create a REST API for user management
/orchestrator Build a responsive e-commerce website
/plan Design a database schema for a blog
/review Check all authentication controllers
/test Run all tests in the current project
/deploy Deploy the application to production

The /orchestrator command is the main entry point. It activates the full 24-agent team with the complete professional workflow:

1. CLARIFY → Asks targeted questions if anything is ambiguous
2. DISCOVER → Scans your project, detects tech stack, reads configs
3. PLAN → Decomposes the mission into milestones with parallel execution groups
4. EXECUTE → Launches specialized agents simultaneously for max throughput
5. VERIFY → Reviewer + QA confirm everything works, zero regressions
6. DELIVER → Summary of what was built, changed, and evidence

---

🌟 Key Features

Feature	Description
24 Specialized Agents	Commander, Planner, Frontend/Backend Devs, Reviewer, QA, PM, DevOps, Security, and more
82 Professional Skills	TDD, security audit, anti-pattern detection, design system, SQL best practices, Docker, Kubernetes, and 75 more
6 Slash Commands	/orchestrator, /orchestrate, /plan, /review, /test, /deploy
Ask Before Building	Agents ask clarifying questions via AskUserQuestion before writing code
Parallel Execution	Commander delegates tasks to multiple agents simultaneously
Persistent Memory	Knowledge Graph via MCP Memory Server stores decisions across sessions
Zero Model Lock-In	Works with Qwen, DeepSeek, OpenAI, Anthropic, or any local model
Multi-Language	TypeScript, PHP (Laravel), Python (Django), Dart (Flutter), Rust, Go, Java, C#
Loop Protection	Monitor agent detects and breaks LLM infinite loops automatically
Quality Gates	Reviewer is the ONLY agent that can approve tasks
Security First	OWASP Top 10 compliance via security-audit skill and Reviewer agent
No Mockups	Agents are forbidden from writing stubs or placeholders
Compaction Recovery	State persists to .qwen-orchestrator/ — survives context window compression

---

👥 Agent Team (24 Specialized Agents)

#	Agent	Role	Superpower
1	Commander 🔴	Master Orchestrator	Parallel delegation, relentless execution
2	Planner 🔵	Research & Architecture	File-level planning, design decisions
3	Frontend Developer 🔵	UI/UX Implementation	Components, responsive, accessible, fast
4	Backend Developer 🟢	Server-Side Logic	APIs, auth, caching, database ops
5	Reviewer 🟣	Quality Gatekeeper	ONLY agent that can approve tasks
6	QA Engineer 🟠	Quality Assurance	Test strategy, edge case discovery
7	Project Manager 🔵	Delivery Management	Scope control, risk assessment
8	Doc Researcher 🟣	Context7 Knowledge	Live doc lookup, anti-hallucination
9	Tech Lead 🟡	Standards & Guidance	Module completeness, CRUD verification
10	Database Architect 🟢	Data Layer Specialist	Schema design, migration safety
11	Product Owner 🟡	Business Value	User stories, acceptance criteria
12	DevOps Engineer 🔘	Infrastructure	CI/CD, Docker, deployment automation
13	Code Quality Guard 🌹	Quality Sentinel	Syntax check, lint, typecheck
14	Monitor 🛡️	Loop Guardian	Detect/break LLM loops, runtime watchdog
15	SEO Specialist 🔵	SEO & Web Performance	Meta tags, structured data, Core Web Vitals
16	Tech Selector 🟣	Technology Advisor	Framework/DB selection with pros/cons
17	Cybersecurity Eng. 🔴	Application Security	OWASP, threat modeling, dependency audit
18	Performance Eng. ⚡	Speed & Scale	Profiling, query optimization, load testing
19	Release Manager 🏷️	Release & Versioning	SemVer, changelogs, rollback planning
20	API Specialist 🔗	API & Integration	REST/GraphQL, versioning, third-party APIs
21	Mobile Engineer 📱	Mobile Apps	Flutter, React Native, offline-first
22	Localization Eng. 🌐	i18n/L10n	Multi-language, RTL, cultural adaptation
23	Documenter 📄	Technical Writing	README, API docs, ADRs, knowledge base
24	Skill Creator 🛠️	Skill Authoring	Skill creation, documentation, testing

---

🛠️ Skills (82 Professional Skills)

Containerization & Orchestration

Skill	Purpose
Docker Containerization	Multi-stage builds, security hardening, image optimization, Docker Compose
Kubernetes Orchestration	Deployments, HPA, service mesh, ingress, GitOps with ArgoCD/Flux

Infrastructure as Code

Skill	Purpose
Terraform IaC	Modules, state management, workspaces, Sentinel policies, Terratest

CI/CD Automation

Skill	Purpose
GitHub Actions CI/CD	Workflows, matrices, environments, caching, composite actions, security

Caching & Performance

Skill	Purpose
Redis Caching	Data structures, caching patterns, session management, pub/sub, Lua scripting

API Design

Skill	Purpose
GraphQL API Design	Schema design, resolvers, DataLoader, pagination, Apollo Federation

Multi-Platform

Skill	Purpose
Multi-Platform Dev	Cross-platform commands, path handling, line endings, permissions, package managers

LLM & Agents

Skill	Purpose
LangGraph	Agent workflows, state management, multi-agent systems
LangChain	LLM chains, agents, tools, memory, RAG
Qwen Agent	Qwen-specific agent development and integration
LLM Integrations	Multi-provider LLM integration (OpenAI, Anthropic, Google, etc.)

Cloud Platforms

Skill	Purpose
Vercel Deployment	Serverless functions, edge computing, Next.js optimization
Cloudflare Pages	Edge functions, global distribution, caching
Cloudflare Workers	Edge computing, request/response manipulation
AWS Serverless	Lambda, API Gateway, SAM, event-driven architectures

SEO

Skill	Purpose
SEO LLM	Optimization for ChatGPT, Perplexity, Gemini, Claude, Bing AI, Qwen
Technical SEO	Site speed, mobile optimization, structured data, Core Web Vitals

Leadership & Architecture

Skill	Purpose
Strategic Leadership	Team management, decision-making, stakeholder alignment
Requirements Engineering	Use cases, user stories, acceptance criteria
Microservices Architecture	Domain-driven design, service boundaries, event-driven communication
Technology Evaluation	Framework/DB selection, pros/cons analysis

Product & Team

Skill	Purpose
Product Owner	Backlog management, user stories, sprint planning
Scrum Master	Ceremony facilitation, team coaching, impediment removal

Software Development

Skill	Purpose
Testing Strategy	Test pyramid, coverage targets, automation
Security Auditor	OWASP Top 10, vulnerability assessment, compliance
DevOps Pipeline	CI/CD design, deployment strategies, IaC
Architecture Patterns	Layered, DDD, microservices, event-driven, CQRS
Documentation	README, API docs, ADRs, knowledge base
Debugging	Systematic investigation, hypothesis-driven debugging
Refactoring	Code smell detection, safe transformations

Security & Quality

Skill	Purpose
Security Code Review	OWASP Top 10, vulnerability patterns, secure coding
Threat Modeling	STRIDE, DREAD, attack trees, threat identification
Test Automation	Framework design, page object patterns, parallel execution
Load Testing	Load, stress, endurance, spike testing
Code Review	OWASP + SOLID + Clean Code systematic review

Database & Data

Skill	Purpose
Database Design	Schema design, relationships, indexing strategies
Database Security	Encryption, access control, audit logging, compliance
SQL Best Practices	Query optimization, indexing, N+1 prevention
SQL Query Assistant	Query assistance, best practices, optimization

Frontend & Mobile

Skill	Purpose
Accessibility	WCAG 2.1 AA, semantic HTML, ARIA, keyboard navigation
Design System	Color palettes, typography, spacing, component architecture
Website Redesign	URL/screenshot-based redesign, analysis workflow
Flutter Web	Responsive layouts, adaptive widgets, state management
Mobile Performance	App optimization, startup time, memory usage
Offline-First	Local storage, sync, conflict resolution

Backend & APIs

Skill	Purpose
API Design	RESTful standards, response envelopes, pagination
API Documentation	OpenAPI specification, Swagger, GraphQL schemas
Third-Party Integration	Stripe, PayPal, AWS, Google, Twilio
Laravel	Eloquent ORM, Form Requests, queues, API patterns
NestJS	Modules, guards, pipes, interceptors, request lifecycle
Supabase	RLS policies, Edge Functions, auth, auto-generated APIs

DevOps & Infrastructure

Skill	Purpose
Disaster Recovery	Backup strategies, restore procedures, failover, RTO/RPO
Release Workflow	SemVer, changelogs, staging, production, rollback
Git Workflow	Branching strategies, commit conventions, PR templates
Deployment	CI/CD, Docker, release management

LLM-Specific

Skill	Purpose
Anti-Hallucination	Source verification, evidence-based claims, validation
Loop Detection	LLM loop detection, tool call loops, escape routes
Context7 Docs	Live documentation lookup via Context7 MCP
Compaction Recovery	Context preservation across session compaction

Project & Team

Skill	Purpose
Agile Project Management	Scrum ceremonies, backlog grooming, sprint planning
User Story Mapping	Story mapping, epic breakdown, timeline visualization
CRUD Completeness	Create, Read, Update, Delete patterns, validation
Project Conventions	Project structure, coding standards, conventions
ERP/SaaS Organization	Large project organization, module structure

Localization

Skill	Purpose
Translation Workflow	Translation memory, machine translation, TMS
Multi-Language	i18n, l10n patterns, RTL support

Additional Skills

Skill	Purpose
Domain-Driven	DDD patterns, bounded contexts, aggregates, value objects
Performance	Profiling, optimization, benchmarks
TDD Workflow	Test-Driven Development (Red/Green/Refactor)
Security Audit	OWASP Top 10 vulnerability detection
Agent Task Coordinator	Agent coordination, MCP protocol
Skill Creation	Skill authoring, documentation, testing
Multi-Channel Funnels	Marketing funnels, conversion optimization
Architecture	System design, patterns, trade-offs
Code Quality	Linting, formatting, type checking
Monitoring	Observability, logging, alerting
Compliance	GDPR, HIPAA, SOC 2, PCI-DSS
Testing	Unit, integration, E2E testing

---

⌨️ Commands (6 Slash Commands)

Command	Description
/orchestrator [goal] ⭐	Main entry point — Full team with clarity protocol (asks questions first)
/orchestrate [mission]	Direct mission execution — no clarity questions, just builds
/plan [feature]	Create implementation plan only — research + architecture
/review [target]	Comprehensive code review — security, quality, patterns
/test [target]	Execute and analyze test suite — coverage, gaps, edge cases
/deploy [target]	Deploy with pre/post verification — CI/CD, rollback plan

When to use /orchestrator vs /orchestrate?

• /orchestrator — When starting a new feature, project, or complex task. The team will ask clarifying questions first.
• /orchestrate — When you already know exactly what you want and just need it executed fast.

---

🧠 How It Works

Execution Flow

User types /orchestrator "Build a payment system"
                    │
                    ▼
         ┌─────────────────────┐
         │   COMMANDER 🔴      │
         │   Clarifies scope   │
         │   via AskUserQuestion│
         └──────────┬──────────┘
                    │
         ┌──────────▼──────────┐
         │   PLANNER 🔵        │
         │   Analyzes project  │
         │   Creates milestones │
         └──────────┬──────────┘
                    │
     ┌──────────────┼──────────────────┐
     ▼              ▼                  ▼
┌─────────┐  ┌──────────┐  ┌──────────────────┐
│FE DEV   │  │BE DEV    │  │DB ARCHITECT      │
│Task 1   │  │Task 2    │  │Schema + Migration│
└────┬────┘  └────┬─────┘  └────────┬─────────┘
     │            │                 │
     ▼            ▼                 ▼
┌─────────┐  ┌──────────┐  ┌──────────────────┐
│REVIEWER │  │QA ENG    │  │CODE QUALITY GUARD│
│Approves │  │Tests     │  │Lint + Typecheck  │
└────┬────┘  └────┬─────┘  └────────┬─────────┘
     │            │                 │
     └────────────┼─────────────────┘
                  ▼
         ┌─────────────────────┐
         │   COMMANDER 🔴      │
         │   Concludes mission │
         │   Delivers summary  │
         └─────────────────────┘

User Clarity: Ask Before Building

The orchestrator never assumes what you want. When a mission starts, the Commander and Planner agents use the AskUserQuestion tool to clarify requirements before writing code.

Example interaction:

👤 User: /orchestrator Build an e-commerce checkout

🤖 Commander: I need to clarify a few things before planning:

   📋 Payment Provider (Question 1/3)
   Which payment gateway should I integrate?
   ○ Stripe       — Industry standard, great API, supports subscriptions
   ○ PayPal       — Widely trusted, good international support
   ○ MercadoPago  — Best for Latin American markets

   📋 Checkout Flow (Question 2/3)
   What type of checkout experience?
   ○ Single Page  — All steps on one page (simpler, faster)
   ○ Multi-Step   — Separate pages per step (more controlled)

   📋 Guest Checkout (Question 3/3)
   Should users be able to checkout without an account?
   ○ Yes          — Lower friction, higher conversion
   ○ No           — Required account, better for retention

👤 User selects: MercadoPago → Multi-Step → Yes

🤖 Commander: Perfect. Planning checkout with MercadoPago, multi-step flow, guest support.

Agents that ask questions:

Agent	When It Asks
Commander	Before every mission — scope, priorities, constraints
Planner	Before architecture decisions — tech stack, patterns
Product Owner	When defining user stories — acceptance criteria, edge cases
QA Engineer	When designing test strategy — critical paths, thresholds
Project Manager	When scoping — deadlines, risk tolerance, resources
Tech Selector	When tech stack is unspecified — presents frameworks, databases, languages with pros/cons
SEO Specialist	When building web projects — target audience, content type, region

Tip: You can always provide full details upfront and skip questions. The agents only ask when they detect ambiguity.

Architecture Diagram

╔══════════════════════════════════════════════════════════════════════════╗
║                    QWEN ORCHESTRATOR v0.0.6                            ║
║                     by Omar-Obando (GitHub)                            ║
╠══════════════════════════════════════════════════════════════════════════╣
║                                                                        ║
║  ┌─────────────────────────────────────────────────────────────────┐   ║
║  │                      COMMANDER 🔴                               │   ║
║  │       "I orchestrate. I never stop until done."                 │   ║
║  │    EXPLORE → LEARN → ADAPT → ACT → VERIFY → DELIVER            │   ║
║  └──────┬─────────┬─────────┬─────────┬────────────────────────────┘   ║
║         │         │         │         │                                 ║
║  ┌──────▼──┐ ┌────▼──────┐ ┌▼────────┐ ┌▼──────────────┐              ║
║  │ PLANNER │ │FE DEV+BE  │ │REVIEWER │ │ QA ENGINEER   │              ║
║  │  🔵     │ │   🟢🟠    │ │  🟣     │ │    🟠         │              ║
║  │Research │ │ TDD Code  │ │Gatekeep │ │Test Strategy  │              ║
║  │Design   │ │ Deliver   │ │Approve  │ │Coverage       │              ║
║  └─────────┘ └───────────┘ └─────────┘ └───────────────┘              ║
║         │         │         │         │                                 ║
║  ┌──────▼─────────▼─────────▼─────────▼──────────────────────────┐    ║
║  │               PROJECT MANAGER 🔵                               │    ║
║  │         Scope · Risk · Progress · Delivery                     │    ║
║  └────────────────────────────────────────────────────────────────┘    ║
║                                                                        ║
║  ┌──────────────┐ ┌──────────────┐ ┌─────────────────────────────┐    ║
║  │DOC RESEARCHER│ │  TECH LEAD   │ │    DATABASE ARCHITECT       │    ║
║  │  🟣 Magenta  │ │  🟡 Gold    │ │       🟢 Teal              │    ║
║  │ Context7     │ │ CRUD Verify │ │   Schema & Migration        │    ║
║  └──────────────┘ └──────────────┘ └─────────────────────────────┘    ║
║                                                                        ║
║  ┌──────────────┐ ┌──────────────┐ ┌─────────────────────────────┐    ║
║  │PRODUCT OWNER │ │DEVOPS ENGINEER│ │   CODE QUALITY GUARD      │    ║
║  │  🟡 Amber    │ │  🔘 Slate    │ │       🌹 Rose             │    ║
║  │User Stories  │ │  CI/CD       │ │  Syntax · Lint · Types    │    ║
║  └──────────────┘ └──────────────┘ └─────────────────────────────┘    ║
║                                                                        ║
║  Agents: 24 | Skills: 82 | Commands: 6 | MCP Tools: 7 | Languages: 8+ ║
╚══════════════════════════════════════════════════════════════════════════╝

---

📦 Installation Guide

Quick Install (Recommended)

qwen extensions install https://github.com/Omar-Obando/qwen-orchestrator

The extension automatically:

• ✅ Registers all 24 specialized agents
• ✅ Loads all 82 professional skills
• ✅ Activates all 6 slash commands
• ✅ Sets up the MCP Memory Server for persistent knowledge
• ✅ Configures everything for immediate use

Manual Install

1. Get the extension:

git clone https://github.com/Omar-Obando/qwen-orchestrator.git

2. Tell Qwen Code about it. Add to your Qwen Code configuration (~/.qwen/settings.json):

{
  "extensions": ["/full/path/to/qwen-orchestrator"]
}

💡 To find your full path, run pwd inside the cloned directory.

MCP Tools Server (Optional)

For advanced features like mission status checking and agent roster tools:

cd qwen-orchestrator
npm install
npm run build

Then add to your Qwen Code MCP configuration:

{
  "mcpServers": {
    "qwen-orchestrator": {
      "command": "node",
      "args": ["./qwen-orchestrator/mcp-server/dist/index.js"]
    }
  }
}

Available MCP tools:

Tool	Purpose
get_mission_status	Get current mission progress
validate_project	Validate project structure and health
generate_todo	Generate structured TODO from mission description
get_agent_roster	List all 24 agents with capabilities
context7_resolve_library	Resolve Context7 library IDs
check_crud_completeness	Verify full CRUD for all entities
create_checkpoint	Create state checkpoint for compaction recovery

Troubleshooting

Problem	Solution
"Command not found" after install	Restart Qwen Code CLI to reload extensions
"Permission denied" on npm commands	Windows: run terminal as Administrator. Mac/Linux: use sudo if needed
Where are session files stored?	.qwen-orchestrator/ directory in your project
Extension not loading	Verify path in settings.json is absolute and correct

Need more help? Open an issue on GitHub.

---

⚙️ Model Configuration

By default, all agents use your Qwen Code default model. There is no hardcoded model — zero lock-in.

⚠️ CRITICAL: Set Temperature to 0.1 (TP 0.1) on ALL models

For the orchestrator to follow instructions strictly and reliably, every model used by any agent MUST be configured with Temperature 0.1 (TP 0.1). Higher temperatures introduce randomness that causes agents to deviate from their assigned roles, ignore constraints, hallucinate tools, or produce inconsistent output.

Why TP 0.1?

• Ensures strict adherence to system prompts and agent instructions
• Eliminates creative drift that breaks multi-agent coordination
• Produces deterministic, reproducible results across sessions
• Prevents agents from inventing tools, files, or behaviors

How to configure:

{
  "modelConfig": {
    "temperature": 0.1
  }
}

This applies to every model — Qwen, DeepSeek, OpenAI, Anthropic, local models, or any other provider. If your provider or model does not support temperature configuration, consider switching to one that does. TP 0.1 is not optional — it is a hard requirement for reliable multi-agent orchestration.

⚠️ CRITICAL: Enable YOLO Mode for Full Agent Autonomy

For agents to operate autonomously without requiring manual approval on every file edit, shell command, or tool invocation, you MUST set Qwen Code CLI to YOLO mode. Without it, every agent action will prompt for confirmation, making multi-agent orchestration impractical.

Why YOLO mode?

• Agents execute tasks without interruption — no per-action confirmation dialogs
• Enables true parallel execution across multiple agents simultaneously
• Required for the Commander to delegate and forget without getting stuck on prompts
• Without it, a single agent task can require dozens of manual approvals

How to configure:

# Set YOLO mode in Qwen Code CLI
/yolo

Alternative modes:

Mode	Behavior	Recommended for
YOLO	Full autonomy — no confirmations. Required for orchestrator.	Production use
Auto-Edit	Auto-approves file edits, confirms shell commands.	Development / testing
Default	Confirms every file edit and shell command.	Learning / evaluation
Plan	Read-only — no file writes allowed.	Planning-only sessions

YOLO mode is not optional for production orchestrator use. Without it, you will be prompted to confirm every single tool call across all 24 agents, making the orchestrator unusable at scale.

Using the default model (recommended)

Just install and use. Every agent picks up whatever model you have configured in Qwen Code.

Setting a specific model per agent

Edit the agent's .md file in agents/ and uncomment the model line:

---
name: commander
# ... other fields ...
# model: uncomment below to override the user's default model
model: qwen-max
---

Recommended models per role

Agent	Recommended Model	Why
Commander	qwen-max or qwen-plus	Needs strong reasoning for orchestration
Planner	qwen-max or qwen-plus	Needs strong analysis for architecture
Frontend Developer	qwen3-coder-plus	Optimized for code generation
Backend Developer	qwen3-coder-plus	Optimized for code generation
Reviewer	qwen-max	Needs deep understanding for quality gates
QA Engineer	qwen-plus	Balanced for test strategy
Code Quality Guard	qwen-plus	Fast for syntax/lint checks
All others	User's default	No special needs

Any model supported by your Qwen Code installation works — including DeepSeek, OpenAI, Anthropic, or local models.

---

🎛️ CLI Mode Recommendations

For maximum autonomy, set your Qwen Code CLI mode before running /orchestrator:

CLI Mode	Setting	What Happens
Auto-Edit	auto-edit	Agents edit/write freely. User confirms shell. Recommended.
YOLO	yolo	Full autonomy. No confirmations. Maximum speed.
Default	default	User confirms every edit. Slower but maximum control.
Plan	plan	Planning only — agents can't write files.

Agents intentionally have no approvalMode, runConfig, or background fields — you control autonomy from the CLI, not the extension.

---

🔧 Advanced Tools

Inter-Agent Communication

The Commander can send messages to background agents mid-task:

SendMessage({ task_id: "worker-auth", message: "User clarified: use JWT, not sessions." })

Real-Time Monitoring

The DevOps Engineer can watch long-running processes:

Monitor({ command: "docker compose logs -f app", description: "Watch app logs during deploy" })

Scheduled Tasks

Schedule recurring quality checks and reports:

CronCreate({ cron: "0 6 * * 1-5", prompt: "Run security audit", recurring: true })

Agent-Tool Matrix

Tool	Cmd	Plan	FE	BE	Rev	QA	PM	Doc	TL	DB	PO	Dev	CQG	Mon	SEO	TSel
AskUserQuestion	✅	✅	✅			✅	✅				✅				✅	✅
Agent (sub-agents)	✅	✅
Skill (load skill)			✅
SendMessage	✅	✅												✅
Monitor (watchdog)	✅						✅						✅
TaskStop	✅												✅
CronCreate/List	✅					✅	✅						✅
ExitPlanMode		✅
SaveMemory	✅	✅	✅	✅	✅	✅	✅	✅	✅		✅			✅	✅	✅
Lsp (diagnostics)	✅	✅	✅	✅	✅			✅	✅			✅	✅
WebFetch	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅		✅	✅

---

🔌 Integrations

Context7 (Optional)

Context7 provides real-time, version-specific documentation.

With Context7: Doc Researcher fetches live docs, zero hallucination on API signatures.

Without Context7: Agents fall back to web search and training knowledge — no errors.

To install, add to your Qwen Code MCP configuration:

{
  "mcpServers": {
    "context7": {
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp@latest"]
    }
  }
}

MCP Memory Server

The @modelcontextprotocol/server-memory MCP server provides a Knowledge Graph that persists across sessions.

What it stores:

• Project decisions: Database choice, framework, architecture patterns
• User preferences: CSS framework, naming conventions, dark/light mode
• Architecture records: Component hierarchy, API conventions, deployment targets
• Session continuity: Last task, next step, known issues

How agents use it:

create_entities({
  entities: [{ name: "database", entityType: "decision",
    observations: ["PostgreSQL 16", "RLS enabled", "UUID primary keys"] }]
})

read_graph({})

The Memory MCP server is configured in qwen-extension.json and runs automatically. No additional setup needed.

UI/UX Skills (Recommended)

The Frontend Developer agent can leverage external UI/UX repos for design intelligence:

Skill	Install	What It Provides
UI UX Pro Max	npx uipro-cli init	161 reasoning rules, 67 UI styles, design system generator, domain search
Designer Skills	/plugin marketplace add Owl-Listener/designer-skills	87 skills, 27 commands across 8 design plugins
Taste Skill	npx skills add https://github.com/Leonxlnx/taste-skill	Anti-slop frontend framework — premium layout, typography, motion, spacing

---

✅ Quality Standards

Every line of code produced by the orchestrator meets:

Standard	Requirement
Complexity	≤ 10 cyclomatic per function
Size	≤ 40 lines per function, ≤ 4 parameters
Types	Strict types, no any
Testing	TDD mandatory, 80%+ coverage
Security	OWASP Top 10 compliance
SQL	Uppercase keywords, one column per line, indented JOINs, N+1 prevention
Review	Multi-dimensional code review before merge
Languages	Not TypeScript-only — adapts to project's tech stack

---

📁 Project Structure

qwen-orchestrator/
├── qwen-extension.json       # Extension manifest
├── package.json              # NPM package config
├── AGENTS.md                 # Agent operational rules
├── LICENSE                   # MIT License
├── CHANGELOG.md              # Release history
├── SECURITY.md               # Security policy
│
├── agents/                   # 24 Agent definitions
│   ├── commander.md
│   ├── planner.md
│   ├── frontend-developer.md
│   ├── backend-developer.md
│   ├── reviewer.md
│   ├── qa-engineer.md
│   ├── project-manager.md
│   ├── doc-researcher.md
│   ├── tech-lead.md
│   ├── database-architect.md
│   ├── product-owner.md
│   ├── devops-engineer.md
│   ├── code-quality-guard.md
│   ├── monitor.md
│   ├── seo-specialist.md
│   ├── tech-selector.md
│   ├── cybersecurity-engineer.md
│   ├── performance-engineer.md
│   ├── release-manager.md
│   ├── api-specialist.md
│   ├── mobile-engineer.md
│   ├── localization-engineer.md
│   ├── documenter.md
│   └── skill-creator.md
│
├── skills/                   # 82 Skill definitions
│   ├── docker-containerization/
│   ├── kubernetes-orchestration/
│   ├── terraform-iac/
│   ├── github-actions-cicd/
│   ├── redis-caching/
│   ├── graphql-api-design/
│   ├── multi-platform-dev/
│   ├── langgraph/
│   ├── langchain/
│   ├── qwen-agent/
│   ├── llm-integrations/
│   ├── vercel-deployment/
│   ├── cloudflare-pages/
│   ├── cloudflare-workers/
│   ├── aws-serverless/
│   ├── seo-llm/
│   ├── technical-seo/
│   ├── strategic-leadership/
│   ├── requirements-engineering/
│   ├── microservices-architecture/
│   ├── technology-evaluation/
│   ├── product-owner/
│   ├── scrum-master/
│   ├── testing-strategy/
│   ├── security-auditor/
│   ├── devops-pipeline/
│   ├── architecture-patterns/
│   ├── documentation/
│   ├── debugging/
│   ├── refactoring/
│   ├── security-code-review/
│   ├── threat-modeling/
│   ├── test-automation/
│   ├── load-testing/
│   ├── code-review/
│   ├── database-design/
│   ├── database-security/
│   ├── sql-best-practices/
│   ├── sql-query-assistant/
│   ├── accessibility/
│   ├── design-system/
│   ├── website-redesign/
│   ├── flutter-web/
│   ├── mobile-performance/
│   ├── offline-first/
│   ├── api-design/
│   ├── api-documentation/
│   ├── third-party-integration/
│   ├── laravel/
│   ├── nestjs/
│   ├── supabase/
│   ├── disaster-recovery/
│   ├── release-workflow/
│   ├── git-workflow/
│   ├── deployment/
│   ├── anti-hallucination/
│   ├── loop-detection/
│   ├── context7-docs/
│   ├── compaction-recovery/
│   ├── agile-project-management/
│   ├── user-story-mapping/
│   ├── crud-completeness/
│   ├── project-conventions/
│   ├── erp-saas-organization/
│   ├── translation-workflow/
│   ├── multi-language/
│   ├── domain-driven/
│   ├── performance/
│   ├── tdd-workflow/
│   ├── security-audit/
│   ├── agent-task-coordinator/
│   ├── skill-creation/
│   ├── multi-channel-funnels/
│   ├── architecture/
│   ├── code-quality/
│   ├── monitoring/
│   ├── compliance/
│   ├── testing/
│   └── debugging-skill/
│
├── commands/                 # 6 Slash commands
│   ├── orchestrator.md
│   ├── orchestrate.md
│   ├── plan.md
│   ├── review.md
│   ├── test.md
│   └── deploy.md
│
├── context/
│   └── QWEN.md
│
├── mcp-server/src/
│   └── index.ts
│
├── docs/
│   ├── ARCHITECTURE.md
│   ├── QUICK-START.md
│   ├── INSTALLATION.md
│   ├── HOOKS-BEST-PRACTICES.md
│   ├── SUBAGENT-MONITORING.md
│   ├── README.es.md
│   ├── README.zh.md
│   ├── README.ja.md
│   ├── README.ko.md
│   ├── README.pt.md
│   ├── README.fr.md
│   └── README.ar.md
│
└── scripts/
    └── setup.sh

---

🎯 First Usage Examples

Once installed, try these in your Qwen Code CLI:

/orchestrator Create a REST API for user management
/orchestrator Build a responsive e-commerce website with Stripe payments
/plan Design a database schema for a blog platform
/review Check all authentication controllers for security issues
/test Run all tests in the current project and report coverage gaps
/deploy Deploy the application to production with rollback plan

---

📚 Documentation

Document	Description
ARCHITECTURE.md	Detailed architecture and design decisions
QUICK-START.md	Getting started guide
INSTALLATION.md	Full installation instructions
HOOKS-BEST-PRACTICES.md	Qwen Code hooks integration guide
SUBAGENT-MONITORING.md	Sub-agent monitoring and lifecycle
QWEN_CODE_PATTERNS.md	MCP servers, hooks, and session management
SECURITY.md	Security policy
CHANGELOG.md	Release history
AGENTS.md	Agent operational rules

---

👤 Author

Omar Obando

• GitHub: @Omar-Obando
• License: MIT

---

🔑 Keywords

qwen-code · qwen-code-extension · multi-agent · ai-agents · orchestration · ai-coding · ai-development · tdd · code-review · devops · cicd · mcp · model-context-protocol · ai-orchestrator · software-engineering · ai-assistant · coding-agent · enterprise-ai · multi-agent-system · ai-coding-assistant · qwen · deepseek · openai · flutter · laravel · nestjs · supabase · typescript · python · php · rust · golang · java · csharp · sql · security-audit · owasp · test-driven-development · code-quality · frontend · backend · mobile · devops · database · api-design · i18n · localization · seo · performance · refactoring · anti-patterns

---