Qwen Code
FreemiumThe open-source AI coding agent that lives in your terminal.
Published 27 September 2026 · Last updated 30 September 2026
Scores
Popularity3/5
27K+ GitHub stars places it among the more established vendor-CLI coding agents, though well behind OpenCode, Cline, and GitHub Copilot in overall community size.
Learning Curve3/5
Terminal-first setup is straightforward for anyone used to a CLI agent, and defaulting to Qwen3-Coder removes an initial model-choice decision, though picking a different provider or configuring MCP servers takes more familiarity.
Flexibility4/5
Accepts any OpenAI-compatible endpoint despite its Qwen-first branding, plus MCP integration and multi-IDE support, giving it real configurability beyond its default model pairing.
Performance4/5
Qwen3-Coder's 69.6% SWE-bench Verified score is genuinely competitive, ahead of the Gemini build the tool was forked from and within range of Claude Sonnet-tier results.
Portability4/5
Apache-2.0 licensed and works with any OpenAI-compatible provider, not just Qwen models, so it avoids hard lock-in even though its documentation and defaults are Qwen-centric.
About Qwen Code
Qwen Code is Alibaba's open-source AI coding agent, built by the Qwen team and released under Apache-2.0. It was originally based on Google's Gemini CLI and has been developed independently since v0.1, growing into a multi-protocol agent that speaks the OpenAI, Anthropic, Gemini, and Qwen APIs. Any other provider works too, as do local models served through Ollama or vLLM, and the model can be switched at runtime. The Qwen models get the deepest integration, but nothing locks you to them.
The agent runs in the terminal as qwen, and is also available as a desktop app for macOS, Windows, and Linux, as plugins for VS Code, Zed, and JetBrains, and through an experimental web UI. Chat channels connect it to Telegram, DingTalk, WeChat, and Feishu. Its agentic toolkit includes Auto-Memory, Auto-Skills, subagents, agent teams, MCP servers, Plan Mode, LSP integration, a sandbox, and git worktrees, while headless mode and TypeScript, Python, and Java SDKs make it scriptable.
It installs with a shell script, npm on Node.js 22 or later, or Homebrew. The /auth menu on first run offers Alibaba Cloud Model Studio, a list of third-party providers, or a custom endpoint. The free Qwen OAuth login ended on April 15, 2026, so hosted Qwen models now run on an Alibaba Cloud subscription or an API key, while local models cost only your own hardware. Qwen Code itself is free, and what you pay for is model access.
Key Features
- Apache-2.0 open-source agent, originally based on Gemini CLI and developed independently since v0.1
- Multi-protocol: OpenAI, Anthropic, Gemini, and Qwen APIs, plus Ollama or vLLM local models, switchable at runtime
- Runs in the terminal, as a desktop app, in VS Code, Zed, and JetBrains, and in an experimental web UI
- Auto-Memory and Auto-Skills retain project context and reusable workflows across sessions
- Subagents, agent teams, Plan Mode, LSP integration, sandboxing, and git worktrees
- MCP server support for extending the agent with external tools and data sources
- Headless mode plus TypeScript, Python, and Java SDKs for scripts and CI
- Chat channels for Telegram, DingTalk, WeChat, and Feishu
Pros
- Apache-2.0 licensed and free to run, with no paid tier of its own and no per-seat fees
- Not tied to Qwen models: any OpenAI-, Anthropic-, or Gemini-compatible provider, or a local Ollama or vLLM model, can be switched in at runtime
- Reaches well beyond the terminal: desktop app, VS Code, Zed, JetBrains, an experimental web UI, and chat apps
- Includes subagents, hooks, MCP, Plan Mode, and skills, a feature set comparable to proprietary terminal agents but with no vendor subscription
- Actively developed, with stable releases landing every few days
Cons
- The free Qwen OAuth login was retired in April 2026, so hosted Qwen models now need an Alibaba Cloud plan or an API key
- Alibaba's Coding Plan key is for interactive use only and may not power automated scripts or application backends, which limits headless and CI runs on it
- Access to Alibaba's plans is restricted: Coding Plan Pro has limited slots restocked daily, and Token Plan is sold only in the Singapore region
- The web UI and daemon mode (`qwen serve`) are still marked experimental
Qwen Code Pricing
Freemium- · Free Apache-2.0 agent for the terminal, desktop, editors, and SDKs; model usage is billed by the provider you connect
- · Works with OpenAI, Anthropic, and Gemini APIs, third-party providers, and local Ollama or vLLM models
- · Personal edition, Singapore region: 11,500 Credits per month for 1-2 concurrent agents
- · Limited-time price; regular price $8/month
- · Personal edition, Singapore region: 25,500 Credits per month for 2-3 concurrent agents
- · Limited-time price; regular price $16/month
- · Personal edition, Singapore region: 45,000 Credits per month for 3-4 concurrent agents
- · Limited-time price; regular price $25/month
- · Alibaba Cloud Model Studio subscription covering Qwen, GLM, Kimi, and MiniMax models
- · Up to 6,000 requests per 5 hours, 45,000 per week, and 90,000 per month
- · Limited slots, restocked daily; interactive use only, no automated scripts
- · Personal edition, Singapore region: 180,000 Credits per month for 6-8 concurrent agents
- · Limited-time price; regular price $80/month
Tech Stacks with Qwen Code
VS Code AI Dev Stack
DeveloperVS Code as the editor, with your choice of AI extension (GitHub Copilot by default, or Cline, Kilo Code, Claude Code, Codex, or Qwen Code) and your pick of GitHub or GitLab for version control.
Development
Qwen Code Dev Stack
DeveloperQwen Code as the coding agent, in the terminal, its desktop app, an editor, or a browser, running Qwen or models such as GLM, Kimi, and DeepSeek, with GitHub or GitLab for version control.
Local LLM Dev Stack
DeveloperA coding agent that works against a model on your own hardware. A local runtime such as Ollama or LM Studio serves an open-weight coding model such as Qwen or Mistral's Devstral, and an agent such as OpenCode or Cline does the editing. No code leaves the machine and there is no per-token bill.
Tools Related to Qwen Code
Integrates with Qwen Code(1)
Qwen Code reaches hosted Qwen models through an Alibaba Cloud Model Studio plan or API key, since the free Qwen OAuth login ended in April 2026, and local Qwen models through Ollama or vLLM.
Alternatives to Qwen Code(12)
Pi is a minimal, model-agnostic terminal harness with four core tools; Qwen Code is a fuller terminal agent tuned for the Qwen models. Pi for building your own agent, Qwen Code for a ready Qwen setup.
Both are open-source vendor terminal agents: Grok Build runs up to eight parallel sub-agents on Grok, Qwen Code is tuned for the Qwen models and accepts any OpenAI-compatible endpoint. Pick by model family and whether you need parallel runs.
Kilo Code is model-agnostic across VS Code, JetBrains, CLI, and cloud; Qwen Code is Alibaba's terminal agent built around the Qwen models. Kilo Code for editor integration and model choice, Qwen Code for a Qwen-tuned CLI.
oh-my-pi is a feature-heavy Pi fork with 31 tools and 60+ model providers; Qwen Code is Alibaba's terminal agent tuned for the Qwen models. oh-my-pi for breadth, Qwen Code for a vendor-optimized open-weight setup.
Qwen Code and OpenCode are both free, open-source, terminal-first coding agents that accept any OpenAI-compatible provider — Qwen Code is built around Alibaba's own Qwen models, while OpenCode has no default vendor.
Qwen Code and ZCode are both agents from a model lab, tuned for its own models (Qwen and GLM). Qwen Code is an Apache-2.0 open-source terminal CLI; ZCode is a desktop app run on a GLM Coding Plan or your own key.
Vendor
Alibaba Cloud
Website →Tags
Details
- License
- Apache-2.0
- Maintained
- Yes
- Primary languages
- JavaScript, TypeScript, Python, Go, Rust, Java
- Type
- AI IDE
- Open source
- Yes
- GitHub stars
- 28.1k
- AI features
- Chat, Agent, Inline edit
- Stars updated
- 2026-09-23