CC Command Center / Local Qwen Advisor — Summary

This batch entry corresponds to the Reddit reference "Claude Code command center Qwen advisor." Two distinct GitHub repositories match this description and are analyzed together:

1. amahpour/claude-code-command-center (5 stars) — A Python/FastAPI web dashboard that hooks into Claude Code's event system to monitor and manage all active sessions in real-time. It installs 8 Claude Code hooks (SessionStart, PreToolUse, PostToolUse, Stop, SubagentStart, SubagentStop, SessionEnd, Notification) via scripts/setup.sh, pipes events through a hook handler to a local SQLite database, and serves a live dashboard at localhost:4700 with session status, cost tracking, live transcripts, full-text search, analytics, and Jira/PR integration.

2. AshlynYang-dev/claude-code-qwen-skill (4 stars) — A Claude Code skill that calls Alibaba's Qwen model (qwen-max/qwen-plus/qwen-turbo/qwen-long) via the Dashscope API as a "local Qwen advisor." It is installed as ~/.claude/skills/qwen, activated with /qwen, and saves responses to markdown files. The skill targets Chinese-language analysis, domestic market research, and second-opinion use cases where Qwen's training data complements Claude's.

Together, these two items describe a workflow where Claude Code sessions are monitored via a local dashboard, and a secondary Qwen model provides specialized advisory input. Neither framework is in the 11 seeds; the closest seed is ccmemory (persistent storage for Claude Code sessions) for the command center component.

CC Command Center / Local Qwen Advisor — Overview

Component A: Claude Code Command Center (amahpour)

Origin

Authored by amahpour, Python/FastAPI, 5 stars, last pushed 2026-05-03. MIT license.

Philosophy

The README states: "A web-based command center for monitoring and managing multiple Claude Code sessions in real-time."

The core value is observability for parallel Claude Code sessions: when running 3-5 agent sessions simultaneously, users lose track of which session is doing what, how much each costs, and whether any need attention. The dashboard provides a single screen for all session status.

Architecture Philosophy (from ARCHITECTURE.md)

"Claude Code has a hooks system — you can register shell commands that run every time Claude does something. We abuse this to spy on every Claude Code session you run, anywhere on your machine, and pipe the data into a web dashboard."

The hook-handler.py bridge is designed to never block Claude Code: zero dependencies (stdlib only), 5-second timeout, silent failure if the server is down.

Component B: Claude Code Qwen Skill (AshlynYang-dev)

Origin

Authored by AshlynYang-dev, Shell/Markdown, 4 stars, last pushed 2026-03-23. MIT license.

Philosophy

"Claude is great at architecture and reasoning. Qwen excels at Chinese-language analysis, domestic market context, and cross-checking data against Chinese real-world norms. This skill bridges both worlds without leaving your terminal."

Use Cases

• Chinese industry research (trained on massive Chinese-language corpus)
• Domestic market benchmarks (restaurant, retail, manufacturing norms)
• Regulatory & compliance questions (Chinese law, standards, business practices)
• Second opinion on numbers (cross-check estimates against Qwen's training data)
• Chinese-language writing (native fluency, not translation)

CC Command Center / Local Qwen Advisor — Architecture

Component A: Command Center Architecture

Distribution

• pip: pip install git+https://github.com/amahpour/claude-code-command-center.git
• Local clone + pip install .

Required Runtime

• Python 3.12+
• Claude Code CLI

Tech Stack

How It Works

Claude Code runs in any terminal
    │
    ▼ (hooks fire on every event)
hook-handler.py (stdlib only, zero deps)
    │ (POST to localhost:4700/api/hooks, 5s timeout, silent fail)
    ▼
FastAPI server (localhost:4700)
    │ (store in SQLite, broadcast via WebSocket)
    ▼
Browser dashboard (real-time updates)

Structure

claude-code-command-center/
├── server/
│   ├── main.py          # FastAPI app
│   ├── sessions.py      # Session state management
│   ├── watcher.py       # JSONL file watcher
│   ├── hooks.py         # Hook event handler
│   ├── db.py            # SQLite + FTS5
│   └── routes/          # API endpoints
├── scripts/
│   ├── setup.sh         # Install hooks into ~/.claude/settings.json
│   ├── hook-handler.py  # Bridge (zero deps, never blocks Claude Code)
│   └── uninstall.sh     # Remove hooks
└── public/              # Dashboard HTML/CSS/JS

Component B: Qwen Skill Architecture

Distribution

git clone https://github.com/AshlynYang-dev/claude-code-qwen-skill ~/.claude/skills/qwen

Required Runtime

• Claude Code
• curl + jq
• Dashscope API key (QWEN_API_KEY)

Structure

claude-code-qwen-skill/
├── SKILL.md           # Claude Code skill definition
├── scripts/
│   └── ask_qwen.sh    # Shell script calling Dashscope API
└── .config.example    # API key template

CC Command Center / Local Qwen Advisor — Components

Component A: Command Center Components

Hooks (8) — installed into ~/.claude/settings.json

All hooks use matcher: "" (match everything, no filtering).

REST API Endpoints

Dashboard Features

Session Status Flow

Working (green) → Waiting (yellow) → Idle (blue) → Stale (grey) → Completed (grey)

Component B: Qwen Skill Components

Skill: qwen

Models Supported

CC Command Center / Local Qwen Advisor — Uniqueness

What Makes This Entry Distinct

This entry covers two small, independently developed repositories that together represent a unique combination not found in any seed framework:

1. Pure observability without workflow prescription (command center)
2. Inter-LLM delegation via skill (Qwen advisor)

Component A: Command Center — Pure Observability Layer

The Gap It Fills

Every seed framework (superpowers, bmad-method, agent-os, taskmaster-ai, etc.) is prescriptive — it tells the agent what to do and how to behave. None of them provide a retroactive observability layer that indexes what the agent actually did across sessions.

The command center is the only tool in the corpus whose sole purpose is observation — no workflow, no methodology, no behavioral prescription. It answers: "What has Claude Code been doing, and what did it produce?"

Searchable Session History

SQLite FTS5 full-text search across all indexed JSONL transcripts is unique in the corpus. Other frameworks with memory (ccmemory, asynkor) store structured state or task records. The command center indexes complete Claude Code transcripts — every assistant turn, every tool call — making them searchable retroactively.

Hook-as-Observer Pattern

The 8-hook configuration covers the full Claude Code lifecycle without modifying any behavior. This is a strict read-only hook usage. In contrast, flokay's hooks validate behavior (write-guard), shep-cli's hooks enforce formatting (PostToolUse → format-tsp.sh), and asynkor's hooks drive coordination. The command center uses hooks purely as a telemetry side channel.

Component B: Qwen Skill — Cross-Model Advisory

The Gap It Fills

No seed framework delegates to a non-Claude LLM from within a Claude Code session. The Qwen skill is the only example in the corpus of:

• Claude Code calling a different vendor's model (Alibaba Qwen via Dashscope)
• A skill acting as a router to a specialized second model
• Cost arbitrage: use a cheaper model (qwen-turbo) for advisory tasks, freeing Claude tokens for implementation

Multi-Model Without Multi-Agent

The Qwen skill achieves a form of multi-model reasoning without multi-agent orchestration. There is no second Claude Code instance, no agent spawning, no coordinator. Claude Code remains the single agent; Qwen is an external oracle it can consult via a synchronous skill call.

This is meaningfully different from codeg (which manages multiple agent instances) or asynkor (which coordinates concurrent Claude Code processes). The Qwen skill is a single-agent, multi-model pattern.

Comparison to Seeds

Limitations

• Low adoption: Combined 9 stars across both repos (5 + 4). Both are proof-of-concept quality.
• No integration: The two components are not integrated with each other. A Qwen skill call does not appear in the command center dashboard (unless the ask_qwen.sh shell execution is captured by the Bash hook).
• Stateless Qwen: No cross-session Qwen memory. Each call is independent.
• Dashboard is read-only: The command center cannot trigger actions, modify sessions, or control Claude Code behavior.