MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework

Sirui Hong,Mingchen Zhuge,Jonathan Chen,Xiawu Zheng,Yuheng Cheng,Ceyao Zhang,Jinlin Wang,Zili Wang,Steven Ka Shing Yau,Zijuan Lin,Liyang Zhou,Chenyu Ran,Lingfeng Xiao,Chenglin Wu,Jürgen Schmidhuber

2023-11-07

Abstract:Remarkable progress has been made on automated problem solving through societies of agents based on large language models (LLMs). Existing LLM-based multi-agent systems can already solve simple dialogue tasks. Solutions to more complex tasks, however, are complicated through logic inconsistencies due to cascading hallucinations caused by naively chaining LLMs. Here we introduce MetaGPT, an innovative meta-programming framework incorporating efficient human workflows into LLM-based multi-agent collaborations. MetaGPT encodes Standardized Operating Procedures (SOPs) into prompt sequences for more streamlined workflows, thus allowing agents with human-like domain expertise to verify intermediate results and reduce errors. MetaGPT utilizes an assembly line paradigm to assign diverse roles to various agents, efficiently breaking down complex tasks into subtasks involving many agents working together. On collaborative software engineering benchmarks, MetaGPT generates more coherent solutions than previous chat-based multi-agent systems. Our project can be found at <a class="link-external link-https" href="https://github.com/geekan/MetaGPT" rel="external noopener nofollow">this https URL</a>

Artificial Intelligence,Multiagent Systems

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve

This paper introduces a new framework called MetaGPT, which aims to enhance the collaboration capabilities of multi-agent systems based on large language models (LLMs) through standardized operating procedures (SOPs). Specifically:

1. **Addressing Issues in Existing Systems**: Existing LLMs-based multi-agent systems often make errors when handling complex tasks due to logical inconsistencies and hallucinations caused by simple chained LLMs. MetaGPT improves this situation by introducing SOPs.

2. **Enhancing Collaboration Efficiency**: MetaGPT improves task coherence and accuracy by breaking down complex tasks into specific executable steps and assigning them to agents with specific roles.

3. **Improving Code Generation Quality**: In software engineering-related benchmarks, MetaGPT has demonstrated better solutions compared to other chat-based multi-agent systems, particularly in terms of code generation quality.

4. **Automating Software Development Processes**: MetaGPT simulates real-world software company workflows, enabling agents to collaborate and divide tasks like humans, thereby improving overall development efficiency and code quality.

In summary, the main goal of this paper is to address the shortcomings of existing LLMs multi-agent systems in handling complex tasks, enhancing collaboration efficiency, and improving code generation quality through the MetaGPT framework.