Synthesis‐GPT: An Intelligent Assistant for Solid‐State Electrolyte Synthesis

All‐solid‐state batteries (ASSBs) promise significant advances in energy density and safety, yet the rapid expansion of literature and heterogeneity of synthesis protocols impede efficient knowledge integration and reproducibility. We present Synthesis‐GPT, a retrieval‐augmented, multiagent large language model (LLM) system that operates over a curated text corpus to extract stoichiometrically accurate synthesis routes and key performance descriptors with full provenance and to organize them into a structured, extensible knowledge base. The system supports two expert workflows—domain question answering and synthesis route extraction—augmented by standardized visualizations (flowcharts and parameter tables) that enable rapid comparison and experiment planning. We demonstrate interactive use cases spanning material recommendation, property lookup, and recipe‐style guidance for high‐conductivity electrolytes. Beyond point queries, Synthesis‐GPT constructs a method‐structured database (solid‐state, liquid‐phase, ultrafast, and other) that supports cross‐route analytics and downstream modeling. By grounding generation in retrieval and coordinating specialized agents, the framework reduces hallucination, preserves domain fidelity, and improves auditability—providing practical infrastructure toward reproducible, closed‐loop, automated materials discovery.