"author:liftkkkk" matching MCP servers:

Exposes deterministic, offline classical NLP capabilities — tokenization, POS tagging, NER, keyword extraction, sentence splitting, OpenIE triples, and sememe/WordNet lookup — as capability-level MCP tools that any LLM agent host can call. Every call returns a unified envelope with exact character offsets, honest engine attribution, and reproducible results, enabling cheap preprocessing, citation-grade spans, and two-stage extraction funnels that keep expensive model reasoning to a minimum.