Zhongke Wenge's ScienceOra research agents go live inside Tencent WorkBuddy

On October 26, Chinese decision-intelligence company Zhongke Wenge put its Panshi ScienceOra research platform live inside Tencent WorkBuddy, releasing more than 30 research-expert agents and skills in the first batch.

The agents cover literature analysis, hypothesis generation, study design, modeling and simulation, validation, and research communication. Users reach them through WorkBuddy's "Discover Apps" entry, which gives them assistance across the whole arc of a project, from getting oriented in a field to supporting specialist decisions.

The technical base is Panshi, a science foundation model developed under the Chinese Academy of Sciences' overall coordination, in which Zhongke Wenge took part. ScienceOra layers the company's science data, research tooling and decision-AI work on top of that, packaging domain knowledge, tools and agent capabilities into one offering and pushing AI beyond retrieval and text generation toward execution of complex research tasks.

Capabilities are staged in three tiers. The entry tier offers agents for academic literature research, comparative literature analysis, deep paper reading and review-map building, plus skills for literature search, multi-paper comparison and structured close reading, all aimed at mapping a field and spotting research gaps.

The deeper tier targets systematic reviews and meta-analyses, molecular biology and multi-omics work, and drug developability assessment. In drug discovery, researchers can draft experimental designs, molecular modeling and simulation plans around candidate molecules, then explore validation and optimization paths against experimental data. The platform also handles paper polishing, manuscript formatting and research slide generation.

The decision tier adds expert agents for rice breeding, materials science, drug R&D and academic evaluation, including a Yuandao Siyuan breeding scientist, a materials characterization and verification scientist and an alloy thermodynamics and phase-field simulation scientist, alongside general reviewer agents and peer-review simulation skills that assess novelty, methodological logic and experimental design.

Yuandao Siyuan is the most concrete vertical application in the launch. The platform was jointly developed by the Hunan Academy of Agricultural Sciences, the CAS Institute of Automation and Zhongke Wenge, and was released this month at Yuelushan Laboratory in Changsha, named in tribute to Yuan Longping. It offers variety-approval lookups, germplasm search, pedigree tracing and literature sourcing, and connecting it to WorkBuddy widens the reach of that breeding knowledge beyond a specialist platform.

Zhongke Wenge was founded by a scientist team from the CAS Institute of Automation and listed on the Hong Kong Stock Exchange in June 2026 under the ticker 1956.HK, with Chinese media describing it as the first listed "decision large model" company. The Panshi line and ScienceOra are its main route for moving decision AI from general question answering into specialist research work.

The distribution logic matters more than any single agent. Putting research agents inside a general-purpose workplace platform is one practical way for vertical AI products to reach users. What to watch is how deeply these agents are actually invoked in real research work, how well the underlying scientific data is maintained and traceable, and how open the platform ecosystem becomes to third-party specialist apps.