MetaGPT Framework
Jump to navigation
Jump to search
A MetaGPT Framework is a role-based multi-agent programming framework by DeepWisdom AI.
- AKA: MetaGPT, Software Company Simulation Framework.
- Context:
- It can (typically) simulate software company workflows through role-based agents.
- It can (typically) implement standardized operating procedures via prompt sequences.
- It can (typically) manage agent roles like product manager, architect, and engineer.
- It can (typically) coordinate multi-agent collaboration through shared environments.
- ...
- It can (often) generate software artifacts from requirement text.
- It can (often) maintain global memory for agent communication.
- It can (often) verify intermediate results through domain expert agents.
- It can (often) support human-ai collaboration via interaction protocols.
- ...
- It can have Agent Roles including product managers, architects, and engineers.
- It can provide Operating Procedures through prompt sequences.
- It can implement Assembly Line Workflows for task decomposition.
- It can maintain Global Memory Pool for collaboration records.
- ...
- Examples:
- MetaGPT Versions, such as:
- MetaGPT v0.1 (2023), introducing initial framework capabilities
- MetaGPT v0.5 (2024), enhancing role-based workflows
- ...
- MetaGPT Applications, such as:
- ...
- MetaGPT Versions, such as:
- Counter-Examples:
- LangGraph Framework, which uses graph-based workflows instead of role-based workflows.
- AutoGen Framework, which employs conversation-based orchestration rather than role-based orchestration.
- CrewAI Framework, which focuses on task delegation over role simulation.
- See: Multi-Agent Framework, Software Development Framework, Role-Based System, Agent Orchestration, Collaborative AI, Software Company Simulation.