China AI Hub
Agents / Qwen-Agent
Qwen-Agent
Alibaba Qwen team's open-source Python framework for developing LLM applications based on Qwen's instruction following, tool usage, planning and memory capabilities (Apache-2.0). Serves as the backend of Qwen Chat (chat.qwen.ai). Ships example applications including BrowserQwen browser assistant, Docker-isolated Code Interpreter, RAG over 1M-token documents, MCP integration and Gradio GUI.
Qwen-Agent is a framework agent by alibaba-cloud, deployable self-hosted, and released open source.
| Key fact | Value |
|---|---|
| Company | alibaba-cloud |
| Type | framework |
| Framework | Python framework (pip install qwen-agent); built-in Assistant / FnCallAgent / ReActChat agents with @register_tool; connects to DashScope API or self-hosted models via vLLM/Ollama |
| Deployment | self_hosted |
| Open source | Yes (Apache-2.0) |
| GitHub | https://github.com/QwenLM/Qwen-Agent |
| Documentation | https://qwenlm.github.io/Qwen-Agent/en/guide/ |
| Pricing | Framework free and open source (Apache-2.0). Model usage billed via DashScope API pay-as-you-go per token, or free with self-hosted open models. No subscription of its own. |
Capabilities
| Capability | Supported |
|---|---|
| Tool calling | Yes |
| Browser use | Yes |
| Computer use | Unknown |
| MCP support | Yes |
| Memory | Yes |
| Planning | Yes |
| Multi-agent | Unknown |
| API | Yes |
Use cases
- Custom LLM applications with tool calling
- Browser automation assistant (BrowserQwen)
- RAG over 1M-token documents
- Code interpreter and PDF-reading assistants
- MCP tool integration and agent evaluation via DeepPlanning benchmark
Limitations
- Docker-based code interpreter has only basic sandbox isolation - use with caution in production
- TIR math demo Python executor is not sandboxed (local testing only)
- GUI requires Python 3.10+
- Last GitHub release v0.0.26 on 2025-05-29; repo development cadence has slowed since
Qwen-Agent is the Qwen team’s open-source Python framework for building LLM applications: it ships built-in Assistant, FnCallAgent and ReActChat agents with a @register_tool decorator, and connects to the DashScope API or to self-hosted models via vLLM/Ollama.
The framework is free under Apache-2.0 with no subscription of its own; model usage is billed per token through DashScope, or free with self-hosted open models.
See the Alibaba Cloud profile.
Sources
Data interpretation
| Field | Value | Evidence type |
|---|---|---|
| Company | alibaba-cloud | Official |
| Type | framework | Official |
| Open source | Yes (Apache-2.0) | Official |
| Deployment | self_hosted | Official |
| Capabilities | Tool calling, Browser use, MCP support, Memory, Planning, API | Official |
Evidence types: Official = vendor documentation or official pages. See the sourcing policy.
What is Qwen-Agent?
Alibaba Qwen team's open-source Python framework for developing LLM applications based on Qwen's instruction following, tool usage, planning and memory capabilities (Apache-2.0). Serves as the backend of Qwen Chat (chat.qwen.ai). Ships example applications including BrowserQwen browser assistant, Docker-isolated Code Interpreter, RAG over 1M-token documents, MCP integration and Gradio GUI.
Is Qwen-Agent open source?
Yes — licensed under Apache-2.0.
How much does Qwen-Agent cost?
Framework free and open source (Apache-2.0). Model usage billed via DashScope API pay-as-you-go per token, or free with self-hosted open models. No subscription of its own.
Where does China AI Hub get its Qwen-Agent data?
From 2 sources, last verified 2026-09-20.