Our take
- Activity
- 19
- Community
- 63
- Issues
- 36
- Pull requests
- 20
+82 stars in the last 30 days
- Slower to close issues than 86% of the projects we track
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
17,268Forks
1,769Change since Nov 16, 2026: +5 stars·0 forks
Overview
Qwen-Agent is a Python framework for building applications around Qwen models. It provides reusable components for model access, tools, and agents, helping developers connect LLM responses to actions such as function calls, document retrieval, and code execution.
It is suited to developers who want an agent framework plus runnable application examples, rather than a standalone chat interface. Models can be accessed through DashScope or a compatible model service, and the repository includes examples such as a browser assistant and code interpreter.
Key Features
- Agent abstractions and built-in assistants for creating tool-using chat applications.
- Function calling support, including parallel tool calls.
- Custom tools can be registered alongside built-in tools.
- MCP integration for connecting agents to configured external tool servers.
- Retrieval-augmented generation examples for document question answering.
- Optional Gradio-based GUI for trying an agent through a web interface.
- Code interpreter tool that runs code in a Docker container.
- Examples for browser assistance and Qwen model tool use.
Use Cases
- Application developers can prototype assistants that call internal or third-party tools in response to user requests.
- Teams working with documents can build question-answering workflows using the provided RAG components and examples.
- Developers exploring MCP can connect agents to MCP servers and expose those tools in an assistant.
- Qwen model users can create interactive demos or applications using DashScope or a compatible model endpoint.
Getting Started
Install the package with optional GUI, RAG, code interpreter, and MCP dependencies:
pip install -U "qwen-agent[gui,rag,code_interpreter,mcp]"
You will also need access to a model service. See the Qwen-Agent README for setup instructions and examples.
Alternatives
- griptape: Griptape is a model-agnostic Python framework for tasks, pipelines, and workflows, rather than a framework centered on Qwen assistants.
- mcp-agent: mcp-agent focuses on MCP-based integrations and composable workflows, including durable execution, rather than Qwen-specific assistant features.
- LazyLLM: LazyLLM emphasizes low-code construction and iteration of multi-agent applications, rather than Qwen-Agent's assistant and tool-calling framework.
- Cheshire Cat AI Core: Cheshire Cat packages customizable agents as API-first microservices with built-in RAG, rather than focusing on Qwen-based assistants.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| Qwen-Agent | Python | Apache-2.0 | 17.1k | Not checked yet |
| griptape | Python | Apache-2.0 | 2.6k | Active |
| mcp-agent | Python | Apache-2.0 | 8.6k | Quiet |
| LazyLLM | Python | Apache-2.0 | 3.9k | Active |
| core | Python | GPL-3.0 | 3.1k | Active |
Considerations
- A model service is required. The README describes DashScope and compatible self-hosted endpoints; DashScope use requires an API key.
- The code interpreter requires Docker. The project warns that its container provides basic isolation and should be used cautiously in production.
- GUI, RAG, code interpreter, and MCP dependencies are optional installation extras.
- The repository notes that its Python executor example is not sandboxed and is intended for local testing, not production use.
Found this useful?
Share it with someone who would like Qwen-Agent.