{"id":5204,"date":"2026-10-29T10:00:00","date_gmt":"2026-10-29T01:00:00","guid":{"rendered":"https:\/\/cinnamon.ai\/?post_type=news&p=5204"},"modified":"2026-10-28T12:49:31","modified_gmt":"2026-10-28T03:49:31","slug":"20261029-super-rag-update","status":"publish","type":"news","link":"https:\/\/cinnamon.ai\/en\/news\/20261029-super-rag-update\/","title":{"rendered":"Enhanced functionality for "Super RAG\u2122" \u2013 Launch of "Starter Plan" with initial setup cost starting from 700,000 yen (excluding tax)"},"content":{"rendered":"

\u2014 Supports Box integration* and web search mode; improved reading of documents including tables and figures.
\u2014 Furthermore, a beta version of the MCP integration will be provided in advance, allowing AI agents to access internal knowledge.<\/p>\n\n\n\n

<\/p>\n\n\n\n

<\/div>\n\n\n\n

Our company will begin offering the latest version (v3.4) of "Super RAG\u2122 (SuperLag)," an enterprise knowledge platform that converts large volumes of complex internal documents into AI-accessible knowledge and provides it to various generating AI and AI agents, from mid-November.<\/p>\n\n\n\n

<\/div>\n\n\n
\n
\"\"<\/figure>\n<\/div>\n\n\n

https:\/\/cinnamon.ai\/super-rag\/<\/a><\/p>\n\n\n\n

<\/div>\n\n\n\n

With "Super RAG" v3.4, we are launching a "Starter Plan" for small-scale deployments, available from an initial setup cost of 700,000 yen (excluding tax). This plan is aimed at companies that want to verify the effectiveness of "Super RAG" in one department or one task before company-wide deployment. In addition, we have added Box integration* and a web search mode, improving the reading and searching of documents including tables and figures. Furthermore, we have expanded the API for connecting with AI agents and are offering a beta version of "MCP (Model Context Protocol)" support, which allows for easy and flexible integration with generative AI services such as Claude and AI assistants such as Microsoft Copilot.<\/p>\n\n\n\n

* Box and SharePoint integration is not available in the Starter Shared Plan and trial environments.<\/p>\n\n\n\n

\u25a0Background and Challenges<\/strong><\/h2>\n\n\n\n

The use of generative AI is expanding from simple question-and-answer sessions to the use of "AI agents" that autonomously support tasks by searching for internal company information. However, accurate reference to internal documents is essential as a prerequisite for integrating AI with business processes. In reality, many companies stumble over the following basic challenges.<\/p>\n\n\n\n

- Unable to extract necessary information from a large volume of internal documents, ranging from thousands to tens of thousands.
- We implemented a generation AI, but we are unable to properly access internal documents.
- Documents containing tables, figures, and scanned PDFs show inconsistent accuracy in responses.
- Internal documents need to be re-registered for each AI service.
- It takes time to verify the basis for the answer.<\/p>\n\n\n\n

Furthermore, new models and services for generative AI and AI agents are constantly emerging, requiring them to be replaced according to the specific business needs. However, there were concerns that if a company's knowledge base had to be re-registered at the same time as switching AI systems, AI utilization would become a "repeated series of individual implementations," leading to inflated operational costs.<\/p>\n\n\n\n

\u25a0 "Super RAG" serves as the foundation for internal knowledge.<\/strong><\/h2>\n\n\n\n

"Super RAG" is designed as an Enterprise AI Knowledge Base that handles the accuracy and quality of extracting enterprise knowledge referenced by the AI, independently of the AI itself, thus smoothly resolving these challenges.<\/p>\n\n\n\n

<\/div>\n\n\n
\n
\"\"<\/figure>\n<\/div>\n\n\n

To ensure accurate referencing of even complex internal documents, our system provides a one-stop solution, from data reading using AI-OCR and structural analysis to highly accurate filtering through hybrid search and reranking, and the addition of context and justification for answers. Even for massive documents spanning thousands to tens of thousands of pages across multiple departments, it generates highly accurate answers* while clearly indicating the source of the references.<\/p>\n\n\n\n

Furthermore, in addition to being usable on the standard chat screen, it also functions as a data platform that integrates with external AI agents, allowing companies to continue using their knowledge as an asset even if they replace the generated AI model in the future.<\/p>\n\n\n\n

*In our verification (conducted in September 2026), while the standard file registration function of typical generation AI tools reached its limit at approximately 350 files, the "Super RAG" integration configuration maintained high answer accuracy even with over 1,000 files registered, and we confirmed that it could correctly present the supporting file names for all questions. Details of the verification can be downloaded from the link below.
"Super RAG" Verification Report Inquiry Form: https:\/\/contents.cinnamon.ai\/download\/superrag_report<\/a>\u3000<\/p>\n\n\n\n

\u25a0 Main contents of "Super RAG" v3.4<\/strong><\/h2>\n\n\n\n

1. Launch of the "Starter Plan" for small-scale businesses.<\/strong><\/h3>\n\n\n\n

For companies looking to start with a single department or task, we offer two plans: "Starter Shared" (initial setup fee 700,000 yen, annual usage fee from 780,000 yen) and "Starter Dedicated" (initial setup fee 2,500,000 yen, annual usage fee from 3,780,000 yen), both providing a shared environment (all prices exclude tax). Both plans allow registration of up to 300 users and offer API and MCP integration, enabling you to connect AI agents, business systems, and internal knowledge from the very beginning.<\/p>\n\n\n\n

*You can upgrade to the Basic plan as your usage expands.
*The Starter plan is available starting with 2 groups and 500 files. The number of groups and available files can be expanded with additional options (for a fee).<\/p>\n\n\n\n

2. Expansion of external knowledge integration<\/strong><\/h3>\n\n\n\n

\u2022 Box integration (new)<\/strong>It supports data connection and automatic synchronization with Box. Knowledge stored in your company's cloud storage can be directly utilized with AI.
- Improved SharePoint integration<\/strong>This feature automatically synchronizes newly added files on SharePoint, significantly reducing the effort required for synchronization setup and operation.
\u2022 Web search mode (new\/optional)<\/strong>External web information can be added to the search scope. This allows information not present in internal knowledge bases to be included in the answers. If the use of web information is not permitted by internal regulations, you can disable it by not selecting the option during implementation (it is required for Starter Shared).<\/p>\n\n\n\n

*Box and SharePoint integration are not available in the Starter Shared Plan and trial environments.<\/p>\n\n\n\n

3. Improved accuracy in reading and searching documents containing tables and figures.<\/strong><\/h3>\n\n\n\n

- Supports page-level chunking of PDF and Excel files, allowing you to search while preserving the information and context within each page.
- "Extended chunking" allows the context surrounding the search results to be included in the response generation process. This helps to reduce incorrect responses caused by missing context.
- The system analyzes the characteristics of tables such as score tables and, if necessary, provides them to the AI as diagrams. This prevents information loss due to the conversion of images into text.
- Improved the attachment upload process and parsing logic.<\/p>\n\n\n\n

4. AI agent integration<\/strong><\/h3>\n\n\n\n

API extension<\/strong>By adding new endpoints and improving existing APIs, we will enhance integration with AI agent development platforms such as Dify and external systems.
\u2022 MCP integration (beta version)<\/strong>We will be providing a beta version of the feature that supports "MCP (Model Context Protocol)," a standard for connecting AI agents and external tools. You can search and refer to "Super RAG" knowledge using standard methods from tools such as Claude and Microsoft Copilot.<\/p>\n\n\n\n

5. Improving stability and operability for large-scale operations<\/strong><\/h3>\n\n\n\n

- Support for handling large amounts of files<\/strong>To ensure stable operation even in knowledge environments with tens of thousands of files, monitoring, alerting, and control functions have been enhanced to suppress unexpected instability.
Token\/Quota Management<\/strong>By managing user tokens and quotas, we enable visibility into usage and cost control.
- User role change function<\/strong>This allows you to change user roles, reducing the effort required for management, verification, and operation.
- Improvement of the administration panel<\/strong>The maximum number of characters for filenames has been increased, and the search function has been improved.<\/p>\n\n\n\n

\u25a0Pricing Plans<\/strong>\u2605Starter plan to be made available with this announcement<\/h2>\n\n\n
\n
\"\"<\/figure>\n<\/div>\n\n\n

Pricing plan for "Super RAG" v3.4<\/p>\n\n\n\n

*All prices are exclusive of tax. The contract period is one year. The annual usage fee is an estimate and includes query usage fees equivalent to 24 million tokens per year (approximately 4,000 question-answer sessions). Web search mode is optional (required for Starter Shared).<\/p>\n\n\n\n

\u25a0 About the "Super RAG" trial<\/strong><\/h2>\n\n\n\n

You can test the search and answer accuracy of "Super RAG" simply by uploading your documents. Companies interested in this service are welcome to contact us. https:\/\/contents.cinnamon.ai\/contact\/inquiry_super-rag<\/a>  Please contact us via the link provided.<\/p>\n\n\n\n

\u2022 Free trial: 5 business days, 1 user, up to 1,000 pages
\u2022 Paid trial: 1 month or more, 700,000 yen (excluding tax), maximum 10 users, up to 5,000 pages (implementation support included)<\/p>\n\n\n\n

*Box and SharePoint integration are not available in the Starter Shared Plan and trial environments.<\/p>\n\n\n\n

\u25a0 Typical Use Cases for "Super RAG"<\/strong><\/h2>\n\n\n\n

\u2022 Manufacturing: Cross-search of lengthy product specifications and technical documents.
\u2022 Finance and Accounting: Cross-sectional survey and comparison of financial statements and related documents
\u2022 Energy Infrastructure: Knowledge creation of equipment and plant-related documents
\u2022 Quality control: Search for QMS, regulations, and quality documents
\u2022 Legal and compliance: Cross-sectional review of contracts, regulations, etc.
Back office: Review and compare forms and non-standard documents.
- Internal inquiries: Handling inquiries using internal knowledge.
AI Agent: Enhancing business operations by leveraging corporate knowledge<\/p>\n\n\n\n

\u25a0 Introducing "Super RAG Pro" for building AI agents<\/strong><\/h2>\n\n\n\n

"Super RAG Pro" is a customized service based on "Super RAG" that supports everything from building AI agents tailored to business challenges and requirements to designing and developing preprocessing, search pipelines, business logic, and UI (business screens). It provides consistent support from requirements definition for AI agent implementation to PoC (Proof of Concept), system development and implementation, operation, and improvement. Beyond simple information retrieval and answer generation, it enables the use of AI in advanced business processes such as those listed below.<\/p>\n\n\n\n

\u2981 Search for drawings: Search for relevant drawings based on shape, symbols, part numbers, etc.
- Complex review and judgment tasks: Reviewing regulations, judgment criteria, application forms, etc., according to established procedures, and supporting the judgment process.
\u2981 Numerical comparison and analysis: Compare numerical data from multiple periods and documents, and calculate differences and aggregated results.
\u2981 AI Agent Development: We design and develop AI agents tailored to business requirements, supporting everything from information retrieval to analysis, decision-making, and task execution.
- Integration with existing systems: We utilize APIs and MCP (beta version) to enable integration with existing business systems and AI agents.<\/p>\n\n\n\n

Business-specific functions and AI agents are built primarily by integrating with "Super RAG" via APIs and MCPs, enabling flexible expansion while minimizing the impact on standard functions. Tuning and updates are also handled on a case-by-case basis, taking into consideration the impact on the system. Implementation proceeds in the following flow: "Hearing (free) \u2192 Requirements definition and PoC \u2192 Development and implementation \u2192 Operation and improvement," and costs are individually estimated according to the customer's requirements.<\/p>\n\n\n\n

\u25a0 About the future<\/strong><\/h2>\n\n\n\n

Cinnamon AI will continue to enhance the functionality of "Super RAG" as a foundation for improving "the quality of corporate knowledge referenced by AI and the accuracy of finding it." MCP integration is currently being prepared for official release after a beta version was provided in v3.4.<\/p>\n\n\n\n

<\/p>\n\n\n\n

Related links<\/strong><\/p>\n\n\n\n

"1TP1 T" Special Website: https:\/\/cinnamon.ai\/super-rag\/<\/a>
"Super RAG" introductory materials: https:\/\/contents.cinnamon.ai\/download\/wp_superrag_dl_lp_super-rag<\/a>
"Super RAG" Implementation Examples:https:\/\/contents.cinnamon.ai\/download\/wp_superrag_5cases<\/a>\u3000
"Super RAG" Verification Report: https:\/\/contents.cinnamon.ai\/download\/superrag_report<\/a>\u3000
Inquiries from companies regarding "Super RAG":https:\/\/contents.cinnamon.ai\/contact\/inquiry_super-rag<\/a>\u3000<\/p>\n\n\n\n

<\/p>\n\n\n\n

<\/p>\n\n\n\n

* "Super RAG\u2122" as used in this press release is a trademark of Cinnamon Co., Ltd.
*The information contained in this press release is current as of the date of publication. Please note that it may be changed without prior notice.<\/p>\n\n\n\n

<\/p>","protected":false},"excerpt":{"rendered":"

\u2015 Box\u9023\u643a*\u30fbWeb\u691c\u7d22\u30e2\u30fc\u30c9\u306b\u5bfe\u5fdc\u3001\u8868\u30fb\u56f3\u3092\u542b\u3080\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u306e\u8aad\u307f\u53d6\u308a\u3092\u6539\u5584\u2015 \u3055\u3089\u306bMCP\u9023\u643a\u306e\u30d9\u30fc\u30bf\u7248\u3092\u5148\u884c\u63d0\u4f9b\u3001\u5bfe\u5fdc\u3059\u308bAI\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u304b\u3089\u793e\u5185\u30ca\u30ec\u30c3\u30b8\u3092\u53c2\u7167\u53ef\u80fd\u306b \u3000\u5f53\u793e\u306f\u3001\u5927\u91cf\u30fb\u8907\u96d1\u306a\u793e\u5185\u30c9\u30ad\u30e5\u30e1\u30f3\u30c8\u3092A […]<\/p>\n","protected":false},"template":"","news-cat":[27],"news-tag":[],"class_list":["post-5204","news","type-news","status-publish","hentry","news-cat-press"],"acf":[],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t