QwenLM released Qwen Code 1.0 Preview, an open-source AI coding agent framework derived from Google Gemini CLI v0.8.2 but independently developed since v0.1. The agent supports multi-protocol LLM access including OpenAI, Anthropic, Gemini, Qwen APIs, and local models via Ollama or vLLM, with runtime provider switching. It features agentic capabilities such as Auto-Memory, Auto-Skills, SubAgents, Agent Teams, MCP integration, dynamic workflows, plan mode, LSP integration, sandbox execution, Git worktrees, and computer use for desktop automation. Interfaces span terminal, VS Code, Zed, JetBrains plugins, a cross-platform desktop app, experimental Web UI via qwen serve --open, and chat integrations for Telegram, DingTalk, WeChat, and Feishu. Automation is enabled through headless mode, TypeScript, Python, and Java SDKs, and a daemon mode providing HTTP+SSE API, multi-workspace support, and session sharing at http://127.0.0.1:4170 with optional bearer authentication. On SWE-bench Verified, seven versions averaged 76.40%–77.80% using Qwen 3.7 Max across 500 tasks with three trials each. The project includes community extensions like Qwen Code Claw for cross-agent delegation via ACP, Aliyun Model Studio CLI, and voice frontend qwen-audio-agent.