AiiDA Copilot: AiiDA Natural Language Interface Using Multi-Agent AI

AiiDA workflows currently require manual Python scripting. This project builds a multi-agent AI system that lets scientists interact with AiiDA through natural language. Five specialized agents handle workflow submission, configuration, diagnostics, analysis, and result queries, connected to AiiDA's Python API via MCP tools. A Validator layer prevents hallucinated parameters from reaching the compute layer. RAG-based retrieval grounds agents in AiiDA documentation. Deliverables: working multi-agent system, MCP server with 5 core tools, RAG pipeline, and architecture documentation.