Qwen Code
Alibaba (Qwen team) · Qwen Code: Open-Source AI Coding Agent for Terminal, Desktop, and IDEs
Qwen Code is the Qwen team's free, open-source (Apache 2.0) AI coding agent. It runs in the terminal, a desktop app, IDEs, and chat apps, and works with Qwen models on Alibaba Cloud or any OpenAI-, Anthropic-, or Gemini-compatible provider. It suits developers who want a Claude Code-style agent without being tied to one model vendor.
Best for
Developers and teams who want a free, open-source, Claude Code-style coding agent that works with Qwen models or any major provider, especially teams already on Alibaba Cloud or using DingTalk, Feishu, or WeChat
Not ideal for
People who want free model access included (the free Qwen OAuth tier has ended), teams that want a managed cloud agent, and anyone who needs a slow-moving tool with long stable release cycles
Who it's for
Developers and engineering teams who want an open-source coding agent with a choice of model providers, including Qwen models on Alibaba Cloud
Capabilities
- Terminal agent (`qwen`) that explains codebases, edits files, runs commands, and fixes bugs
- Qwen Code Desktop app for macOS, Windows, and Linux
- IDE integrations for VS Code, JetBrains, and Zed, plus an experimental browser web UI
- Chat channels for Telegram, DingTalk, WeChat, and Feishu via `qwen channel start`
- Subagents, agent teams, dynamic workflows, auto-memory, auto-skills, and hooks
- MCP support, plan mode, LSP integration, sandbox, git worktrees, and computer use
- Agent Arena for running several models head-to-head on the same task
- Headless mode (`qwen -p`) for scripts and CI, plus TypeScript, Python, and Java SDKs
- Experimental daemon mode (`qwen serve`) that lets multiple clients share one agent over HTTP and SSE
- OpenAI-compatible, Anthropic, and Gemini protocols (including Google Vertex AI), with runtime model switching
Limitations
- The free Qwen OAuth tier was discontinued on April 15, 2026, so you need an Alibaba Cloud plan, a third-party API key, or a local model
- The web UI and daemon mode are marked experimental
- The standalone installers download from Alibaba Cloud storage in Hangzhou. npm and Homebrew installs are alternatives
- The Python and Java SDKs are in alpha
- npm installation requires Node.js 22 or later
- Very frequent releases (seven stable v0.24.x versions in ten days in September 2026) mean behavior and settings change often
- No hosted cloud agent is offered yet. It runs on your machine or your own server, although recent release notes show hosted-runtime work in progress
Use cases
- Asking the agent to explain an unfamiliar repository and point out where to start
- Implementing a feature or bug fix from the terminal or Qwen Code Desktop and reviewing the diff
- Running `qwen -p` in CI to triage failing tests or draft fixes
- Comparing how different models handle the same task with Agent Arena
- Running coding tasks from a DingTalk, Feishu, or Telegram chat
- Using a self-hosted model through Ollama or vLLM for code that cannot leave your network
Our take
Qwen Code has grown from a terminal agent into an open-source coding agent with an unusually wide set of surfaces, including a desktop app, IDE integrations, SDKs, and chat channels, while staying provider-neutral. Its main advantage over vendor agents is choice: the same tool can drive Qwen, Claude, Gemini, OpenAI, or a local model. The end of the free Qwen OAuth tier removes what used to be its easiest on-ramp, so it now competes on features and openness rather than free access. The pace of releases is a strength and a cost, and teams should pin versions for anything automated.
Who should use it
Developers who want a full-featured open-source coding agent with a choice of models, teams already using Qwen models or Alibaba Cloud, and teams that work in DingTalk, Feishu, or WeChat and want to trigger coding tasks from chat.
Who should skip it
People looking for free model access included with the agent, teams that need a hosted cloud agent, and organizations that cannot absorb frequent tool updates.
Strengths
- Free and open source under Apache 2.0
- Works with Qwen models or any OpenAI-, Anthropic-, or Gemini-compatible provider, including local models
- Many surfaces: terminal, desktop app, IDEs, web UI, SDKs, and chat channels
- Broad agent feature set, including subagents, MCP, hooks, sandbox, and worktrees
- Very active development
Weaknesses
- Free Qwen OAuth access has ended, so model usage now costs money unless you run a local model
- Web UI and daemon mode are still experimental
- Fast release cadence means frequent changes
- Chat integrations lean toward Chinese platforms (DingTalk, WeChat, Feishu) alongside Telegram
Qwen Code pricing
Open Source
Free
- Apache 2.0 CLI, desktop app, IDE integrations, and SDKs
- Bring your own model provider, Alibaba Cloud plan, or local model
Note: Qwen Code itself is free. Model usage is billed by the provider you connect: Alibaba Cloud's Coding Plan (fixed monthly fee) or Token Plan (billed per token), or another provider's API. The free Qwen OAuth tier ended on April 15, 2026.
Technical specs
API pricing
Free: headless CLI, SDKs, and experimental daemon; model usage is billed by the provider or Alibaba Cloud plan you connect
Available models
Where Qwen Code excels
Provider-neutral coding agent for a team
One tool works with Qwen, OpenAI, Anthropic, Gemini, or self-hosted models, so teams can switch providers without switching workflows.
Model evaluation on real tasks
Agent Arena runs several models against the same task, which helps teams decide which model to pay for.
Automation in CI
Headless mode and the TypeScript, Python, and Java SDKs let teams run the agent from scripts and pipelines.
Qwen Code vs. competitors
Qwen Code vs. Claude Code
Claude Code is Anthropic's commercial coding agent for Claude models. Qwen Code offers a similar feature set as open source and works with many providers, including Anthropic models through its Anthropic protocol support.
Qwen Code vs. OpenCode
OpenCode is an open-source, multi-provider terminal coding agent. Qwen Code is also open source and multi-provider, and adds chat channels, Agent Arena, and first-party ties to Qwen models and Alibaba Cloud plans.
Qwen Code vs. Kimi Code
Kimi Code is Moonshot AI's coding agent built around Kimi models. Qwen Code is the Qwen team's equivalent, with an Apache 2.0 codebase that also connects to OpenAI, Anthropic, Gemini, and local providers.
Qwen Code vs. goose
goose is a foundation-governed, open-source general-purpose agent with recipes and cron scheduling. Qwen Code focuses on coding, with IDE integrations, LSP, git worktrees, and a Claude Code-like command set.
Frequently asked questions
What is Qwen Code?
Qwen Code is an open-source AI coding agent from Alibaba's Qwen team. It runs in your terminal, a desktop app, IDEs, a web UI, or chat apps, and it reads code, edits files, runs commands, and automates development tasks.
Is Qwen Code free?
The software is free and open source under Apache 2.0. Model usage is not free by default: the free Qwen OAuth tier was discontinued on April 15, 2026, so you use an Alibaba Cloud Coding Plan or Token Plan, an API key from another provider, or a local model.
Which models does Qwen Code support?
Qwen Code supports OpenAI-compatible, Anthropic, and Google Gemini protocols, including Google Vertex AI. That covers Qwen models on Alibaba Cloud Model Studio, OpenAI, Anthropic, Google, OpenRouter, ModelScope, Azure, and self-hosted endpoints such as Ollama or vLLM.
Is there a Qwen Code desktop app?
Yes. Qwen Code Desktop is available for macOS, Windows, and Linux from the project's GitHub releases, and it is published alongside CLI releases.
Qwen Code vs Claude Code: what is the difference?
Both are agentic coding tools with subagents, MCP, hooks, plan mode, and IDE integrations. Claude Code is Anthropic's commercial product for Claude models. Qwen Code is Apache 2.0 open source, works with many providers, and adds Agent Arena, a multi-client daemon mode (`qwen serve`), and chat channels for DingTalk, WeChat, and Feishu as well as Telegram.
Integrations & fit
Alternatives to consider
Anthropic
Developers and teams who want deep codebase reasoning, multi-file changes, and persistent project instructions in a terminal-friendly or IDE-native workflow
Paid
Anomaly
Developers who want a terminal-based AI coding agent with no vendor lock-in, the ability to bring any LLM provider, and undo control over agent file changes
Free
Moonshot AI
Developers who want Moonshot's Kimi K3 models in one coding agent across the terminal, a desktop app, VS Code, and ACP editors, paid for through a Kimi membership
Paid
Agentic AI Foundation (Linux Foundation)
Developers and technical teams who want a free, vendor-neutral agent on their own machine, with a desktop app as well as a CLI, MCP extensions, and shareable recipes they can schedule or run in CI
FreeAbout Qwen Code
Qwen Code is an agentic coding tool from Alibaba's Qwen team that reads your codebase, edits files, runs commands, and automates repetitive development work. You start it with `qwen` in a project directory, or use the Qwen Code Desktop app for macOS, Windows, and Linux, the VS Code, JetBrains, and Zed integrations, an experimental web UI (`qwen serve --open`), or chat channels on Telegram, DingTalk, WeChat, and Feishu. The README states that the project has worked toward feature parity with Claude Code, and it lists subagents, agent teams, auto-memory, skills, hooks, MCP, plan mode, LSP integration, a sandbox, git worktrees, and computer use. A few features go beyond that set: Agent Arena runs several models on the same task side by side, and daemon mode (`qwen serve`, experimental) lets multiple clients share one agent over HTTP and SSE. Headless mode (`qwen -p`) and TypeScript, Python, and Java SDKs cover scripting and CI. Qwen Code is not tied to one model vendor. It speaks OpenAI-compatible, Anthropic, and Gemini protocols (including Google Vertex AI), so you can connect OpenAI, Anthropic, Google, OpenRouter, ModelScope, Azure, or a self-hosted endpoint such as Ollama or vLLM. The recommended path is Alibaba Cloud's Coding Plan (fixed monthly fee) or Token Plan (pay per token). The main change for past users is that the free Qwen OAuth tier was discontinued on April 15, 2026, so you now need a paid plan, an API key, or a local model. The project releases very often, with stable versions v0.24.0 through v0.24.6 shipping between September 16 and 26, 2026, so settings and behavior change quickly.
Updates from Qwen Code
The Qwen team shipped Qwen Code v0.24.6 alongside Qwen Code Desktop v0.24.6.
Qwen Code v0.24.0 added dynamic workflows shipped by extensions, shared channel output modes with DingTalk support, and per-session caps on web search calls.
The free Qwen OAuth tier was discontinued. Users now authenticate with an Alibaba Cloud Coding Plan or Token Plan, or an API key for another provider.
Are you the founder? Claim this listing →