Qwen Code open-source terminal coding agent released, with multi-protocol and multi-platform support

Qwen Code is an open-source AI coding agent, with both its framework and Qwen models open-sourced. It ships with Auto-Memory, Auto-Skills, SubAgents, Agent Teams, and MCP, and supports OpenAI, Anthropic, Gemini, and Qwen APIs as well as local models via Ollama and vLLM, switchable at runtime. Beyond the terminal, it offers IDE plugins, a desktop app, a Web UI, an SDK, and chat integrations for Telegram, DingTalk, WeChat, and Feishu. Installation options include Linux/macOS scripts, Windows commands, NPM (Node.js 22+ required), and Homebrew. On SWE-bench Verified, seven Qwen Code versions using the same Qwen 3.7 Max model averaged 76.40%–77.80% (500 tasks, three trials per version). The project was initially based on Google Gemini CLI v0.8.2, but stopped syncing upstream from Qwen Code v0.1, evolving independently into a multi-protocol, multi-platform agent framework. Claude Code users may want to note Qwen Code's feature parity plus extras like multi-protocol support, Daemon mode, and IM channels; teams running local open-source models can connect via Ollama or vLLM.