PowerWorkflow is an open-source collection of agentic workflows for power system applications. These workflows enable intelligent automation and coordination of power system operations, facilitating efficient management and control in the energy domain.
Agentic workflows are step-by-step processes that process logic together. They provide a structured way to implement intelligent behavior without explicit agent programming. Key features include:
- Automated execution of complex power system operations
- Integration with various power system software and tools (PowerMCP)
- Reusable workflow components for different use cases
- Flexible orchestration of multiple tools and services
We're building an open-source community focused on advancing workflow automation in the power domain. Our goals are:
- Collaboration: Connect power system experts, workflow developers, and automation specialists
- Innovation: Create and share powerful workflows for various use cases
- Education: Provide resources and examples for implementing agentic workflows
- Standardization: Develop best practices for workflow design in the energy sector
Check out these demos showcasing PowerWorkflow in action:
- Grid Impact Evaluation for Large Load Interconnection: This demo showcases how Agentic AI can streamline grid impact evaluation for large loads like data centers. Using domain-adapted AI agents equipped with tools such as PowerWorld, the workflow automates data analysis, contingency studies, and report generation.
- Electricity Market Expert Demo: An automated workflow that handles electricity market inquiries, generates professional responses using an LLM, logs actions for traceability, and supports human-in-the-loop review.
- Cross-Engine Transient Simulation Workflow: An agent-facing PowerWF workflow for executing one canonical transient-stability study through two independent simulation MCPs, preserving both outcomes, and delegating deterministic trajectory comparison to PowerMCP. The workflow is engine-agnostic and reports engine/comparator failures explicitly.
- B2G-Agent: Building--Grid Co-Design: An LLM-mediated workflow in which a Building Engineer and a Distribution Power Engineer negotiate grid-upgrade or demand-response decisions through natural language. The workflow translates between disciplines, delegates analysis, preserves an auditable decision record, and keeps engineers responsible for final approval. A runnable source snapshot is included in this repository. B2G-Agent is independently developed and maintained by Xueyuan Cui; ongoing development remains in the canonical cuixueyuan/B2G-Agent repository.
- FlowiseAI: Open-source visual workflow builder
- Dify: AI workflow development platform
- Coze: Bot development platform
- n8n: No-code workflow builder
- Gumloop: No-code workflow builder
- Relevance AI: No-code Agent and workflow builder
- Microsoft Copilot Studio: Enterprise workflow development
- LangGraph: Python framework for building agent workflows
- AutoGen: Microsoft's framework for collaborative agents
- CrewAI: Framework for collaborative AI agents
- A2A Protocol: An application-level protocol that enables agents to collaborate in their natural modalities
We welcome contributions! Please see our Contributing Guidelines for details.
This project is licensed under the MIT License - see the LICENSE file for details.
- Lin Dong, Qian Zhang
- All contributors who help make this project better
- The Power and AI Initiative (PAI) at Harvard SEAS

