qwen-code

qwen-code Plugin

qwen-code (qwen binary) is Alibaba's agentic coding CLI — a Gemini CLI fork tuned for Qwen3-Coder. It drives an agentic loop over your repo: file tools, shell tool, memory, and MCP servers. qwen -p makes it scriptable.

Commands

• qwen-code self version — Print version
• qwen-code agent ask "<prompt>" — Headless one-shot run (qwen -p), works in current dir
• qwen-code agent chat — Interactive TUI (opens the full agent interface)
• qwen-code mcp manage — qwen mcp list|add|remove MCP servers
• qwen-code _ _ — Passthrough (--all-files, --yolo, extensions, ...)

Usage Examples

• "Ask the qwen agent to explain this codebase"
• "Run qwen headlessly to generate tests for src/foo.py"
• "List the MCP servers qwen-code has configured"

Installation

npm install -g @qwen-code/qwen-code   # Node 18+

Auth: Qwen OAuth on first qwen run (free tier) or OPENAI_API_KEY/OPENAI_BASE_URL/OPENAI_MODEL for any OpenAI-compatible endpoint — including a local vllm serve or llama.cpp server.

Examples

# Headless agent task (prints result, exits)
qwen -p "find and fix the bug in src/login.py"

# Auto-approve every tool call (dangerous, sandbox it)
qwen -p "migrate this file to async" --yolo

# Include all files in context
qwen -p "refactor for readability" --all-files

# MCP server management
qwen mcp list
qwen mcp add myserver -- npx -y my-mcp-server

Key Features

• Headless -p/--prompt mode → scriptable agent runs; --yolo for full auto
• Repo-aware agentic loop: read/edit/grep/shell/todo tools out of the box
• MCP server support (qwen mcp add ...), extensions, session checkpointing
• Runs against Qwen OAuth free tier or any OpenAI-compatible endpoint