AutoAgent: An Open-Source Framework for Natural-Language LLM Agents

AutoAgent is an open-source framework from the Data Science and AI Laboratory, The University of Hong Kong (HKU-DSAI Lab).
It was introduced in the paper AutoAgent: A Fully-Automated and Zero-Code Framework for LLM Agents (arXiv:2502.05957) by Jiabin Tang, Tianyu Fan, and Chao Huang.

This project enables researchers and engineers to create and deploy large-language-model (LLM) agents using natural language alone.
By removing the need for manual coding, AutoAgent lowers the barrier for both rapid prototyping and production deployment.

Key Capabilities

• Agentic-RAG with Native Vector Database
  Built-in, self-managing vector storage designed to outperform traditional pipelines such as LangChain.
• Zero-Code Agent and Workflow Design
  Define tools, workflows, and multi-step agents entirely through natural-language prompts—no boilerplate code required.
• Broad LLM Compatibility
  Works with major providers including OpenAI, Anthropic, Deepseek, vLLM and Hugging Face.
• Flexible Reasoning Modes
  Supports both function-calling and ReAct interaction for complex tasks.
• Lightweight and Extensible
  Designed to be dynamic and customisable, making it suitable for both research and real-world applications.

Why It Matters

AutoAgent provides a reproducible, academically grounded platform for developing intelligent agents without proprietary lock-in.
Its natural-language interface accelerates experimentation while maintaining the robustness needed for advanced projects.

Read the full paper for technical details: arXiv:2502.05957

Explore the code and contribute here: https://github.com/HKUDS/AutoAgent