Qwen Code is the Qwen team’s open-source command-line tool, derived from Gemini CLI and developed independently since its first versions. It is a terminal agent, not a model: the entry describes the tool that uses models.
What it’s for
- Running multi-step coding tasks right in the terminal
- Reading and editing project files, running commands and iterating on the result
- Connecting external tools and context through the MCP protocol
- Automating flows in headless mode, with SDKs for common languages
How it works
- You install the terminal command and choose which model provider to use
- The tool speaks multiple API protocols, including formats compatible with OpenAI, Anthropic and Gemini
- Besides Qwen models, it accepts third-party providers and local model execution
- The same core serves interactive use, a web server and editor plugins
Where it runs
- Terminal, as the main command
- Desktop app and a locally served web interface
- Plugins for editors such as VS Code, JetBrains and Zed
- CI pipelines and headless execution
Availability and licensing
- Open-source tool under the Apache-2.0 license
- It does not require a subscription to the tool itself to use models on your own
- Paid access to Qwen models happens through the corresponding cloud platform
How it differs
- It is the tool, not the Qwen model family nor the cloud platform
- It accepts models from several vendors, avoiding dependence on a single provider
- It keeps a historical common base with another CLI but follows independent development
Things to note
- The agent runs real commands; reviewing diffs and permissions is part of the flow
- It requires an environment with a recent JavaScript runtime version
- Consumption and cost vary with the chosen model provider