QwenLM/qwen-code
An open-source AI coding agent that lives in your terminal.
What it solves
Qwen Code is an open-source AI coding agent designed to automate software development tasks directly from the terminal, IDEs, or desktop applications. It removes vendor lock-in by supporting multiple LLM providers and open-source models, allowing developers to switch between different AI models at runtime.
How it works
The tool operates as a multi-protocol agent framework that can be integrated into various environments. It can be run as an interactive terminal UI, a headless mode for CI/CD scripts, a daemon for shared agent sessions, or via SDKs in Python, TypeScript, and Java. It leverages agentic workflows including Auto-Memory, Auto-Skills, and the ability to spawn SubAgents and Agent Teams to handle complex coding tasks.
Who it’s for
It is built for developers who want a powerful, open-source alternative to proprietary coding agents, and those who need deep integration with their existing development environment (terminal, IDEs like VS Code, Zed, and JetBrains).
Highlights
- Multi-protocol support: Compatible with OpenAI, Anthropic, Gemini, and Qwen APIs, as well as local models via Ollama or vLLM.
- Agentic capabilities: Features built-in skills for reviewing, batch processing, and bug fixing, along with support for MCP and LSP integration.
- Extensible interface: Available as a CLI, desktop app, IDE plugins, and IM bots for platforms like Telegram and WeChat.
- Open-source ecosystem: Both the framework and the underlying Qwen models are open-source, allowing the tool to iterate on its own code via AI.
- Agent Arena: Allows head-to-head comparison of multiple models on the same coding task.