Textweaver launches TaggingBox Server, an enterprise AI knowledge platform supporting Korean and HWP documents
Textweaver launches TaggingBox Server, an enterprise AI knowledge platform supporting Korean and HWP documents
Textweaver (텍스트웨이), a South Korean deep-tech startup, has officially launched TaggingBox Server (태깅박스 서버), an enterprise AI knowledge management and reasoning platform that supports Korean-language documents and the domestic HWP format. The company completed proof-of-concept (PoC) deployments with over 30 organizations, including public institutions, daily newspapers, and maritime antenna manufacturers, before the commercial release. Textweaver reported 2025 revenue of 394 million won (~$295,000) from license sales, with a 20%+ conversion rate from PoC to paid contracts. The startup was selected for South Korea's DeepTech TIPS program and the Youth Startup Academy DeepTech track in 2025.
The launch matters because Textweaver is positioning TaggingBox Server as a "white-box AI" alternative to opaque retrieval-augmented generation (RAG) systems, using its proprietary PISA engine to decompose information into predicate-centric atomic event units and reconstruct them as a knowledge graph (PKG) with verifiable reasoning paths (PathHash). This directly addresses the structural trust deficit that prevents many enterprises, especially in regulated sectors, from deploying LLMs for internal knowledge work. By supporting the HWP format — a standard in Korean public-sector and enterprise environments — the company is building a local data-moat that hyperscaler LLM providers cannot easily replicate. The pattern echoes the "context-engineering moat" observed in enterprise AI, where domain-specific document handling and traceable inference become defensible differentiators.
Textweaver is also pursuing global expansion, with a completed PoC and patent filing in Singapore through partner Tictag, and collaboration with San Jose State University. A global launch event is scheduled for June 25 in Brooklyn, New York. The company's emphasis on data sovereignty for Korean enterprises — keeping sensitive documents on-premise while delivering structured, auditable AI reasoning — positions it within the broader sovereign AI trend. However, the modest revenue base and early stage of the PISA engine (slated for full integration in H2 2026) mean the company must prove it can scale beyond small-batch PoCs. If successful, Textweaver could become a case study for how regional document formats and regulatory environments create viable niches for specialized enterprise AI platforms.
#Textweaver #TaggingBox #EnterpriseAI #KnowledgeManagement #KoreanAI #WhiteBoxAI