​

Deskrex Launches Snorbe, a Self-Evolving AI Research Agent

Deskrex Inc. (Headquarters: Shibuya, Tokyo; CEO: Itaru Tomita) announces the general availability of Snorbe’s Closed Alpha version starting May 26, 2026.

​

What is Snorbe?

Snorbe is a self-evolving research AI built on the concept “Research becomes emergence.” It accumulates research results — from academic papers, patents, news, and internal documents — as semantic vector, knowledge graph, and string indexes. Its proprietary exploration algorithm automatically detects unexplored areas (white spaces) within your knowledge. The more you interact with Snorbe, the more it remembers context across sessions. By connecting past research with new findings, Snorbe surfaces perspectives and ideas that even the researcher hadn’t considered.

​

Background

R&D patent and paper surveys, sales planning target discovery, new business idea exploration — these research workflows share common challenges:

• The sheer volume of literature and information sources is overwhelming
• Output quality varies significantly depending on the researcher’s experience and skill
• Past research assets are managed in silos, making it hard to accumulate and share organizational knowledge

Most existing AI research tools are designed as “one-question, one-answer” systems. Deskrex built Snorbe to structure the research process itself, enabling knowledge to accumulate and be leveraged as an organizational intellectual asset.

​

Key Features

​

1. White Space Detection via Knowledge Graph

As research accumulates in the graph, unexplored areas emerge naturally. Community classification automatically groups related information, providing an overview of your entire research landscape.

​

2. Agent Memory That Grows With Every Conversation

Context is remembered and accumulated across sessions every day. Because the agent remembers past industry trends and competitor intelligence, you can give instructions like “Compare this with A Company’s patent portfolio from last time.”

​

3. Reproducible Outputs with Specialized Skills

Snorbe includes “Skills” — Markdown-based workflow instructions based on the format advocated by Anthropic. Official skills cover academic research, patent search & analysis, market & investment data research, and more.

​

4. Multi-Model, Multi-Agent, Multi-Source Support

Compatible with AI models from OpenAI, Anthropic, Google, and xAI. The system automatically selects the optimal model for each stage of research.

​

5. Automation & Team Sharing

Supports daily, weekly, and monthly automated research execution. Knowledge and memory data can be shared across team members.

​

6. External Memory for Agent-to-Agent Integration

Snorbe’s research capabilities can be called from external AI agents such as Claude Code and Codex, acting as “external memory” for your own AI systems.

​

Use Cases

​

Launch Support Campaign

For companies adopting Snorbe during the Closed Alpha period:

1. Custom Skill Development — Deskrex’s team creates and delivers custom skills tailored to your specific workflow
2. Research BPO Service — Deskrex consultants handle patent/paper analysis and business exploration research on your behalf

​

How to Get Started

The Closed Alpha is available free of charge for applicants. Apply via the service page to receive your authentication link and trial credits.