Softcreate Co., Ltd. has added a new method called “MCP Integration” to its AI generation service for companies and organizations, “Safe AI Gateway,” which enables the structural use of kintone data. This will allow for more stable answers to questions such as aggregation and conditional extraction of business data. In addition, a new search mode called “Web Search (Simple Search)” that prioritizes speed and cost has been added to the web search agent, and the image generation model “GPT-Image-2” has been launched, further streamlining information utilization and content creation in business operations.
MCP integration enables aggregation and conditional extraction of business data by understanding the “data structure.”
In the conventional RAG-based kintone integration, information within kintone is searched and referenced as text to generate answers. As a result, answers to data-driven questions, such as tracking the number of items by status or aggregating data based on specific conditions, were sometimes inconsistent. Operational adjustments, such as tuning system prompts, were sometimes necessary to obtain the expected answers.This MCP integration allows us to reference data structures such as kintone records and fields, enabling us to more reliably obtain answers tailored to the specific criteria we want to see, such as “aggregating by status” or “extracting only overdue items.”
In the future, it is envisioned that this will be expanded to include business operations such as updating records through chat-based communication.
Enhanced functionality to further improve operational efficiency (Web search, image generation)
Also Read: AIstudio MiYaGi Debuts Creative AI Video Platform
A new search mode, “Web Search (Simple Search),” to meet the need for “quick searches.”
Safe AI Gateway’s web search agent has traditionally offered a deep-dive research mode called “Web Research (Advanced Search)” that recursively collects information by traversing multiple sites, verifies the evidence, and provides answers in a report format.
This update adds a new “Web Search (Quick Search)” feature that prioritizes speed and allows you to quickly grasp the overview. By gathering information across multiple sites without recursive in-depth analysis, it achieves both quick answers and reduced token consumption.
This makes it possible to use “Web Research (detailed search)” and “Web Search (simple search)” interchangeably depending on the purpose and nature of the work.
Added image generation function “GPT-Image-2”.
This version update incorporates “GPT-Image-2,” the latest image generation model provided by Microsoft’s Azure OpenAI Service.
This model significantly improves practicality compared to conventional image generation AI by considering composition, layout, and text placement before generating images. In particular, the accuracy of text rendering, including Japanese text, has been improved, enabling the generation of images of a quality that can be used directly in banners, documents, and other materials.
This allows image creation, which previously required adjustments and revisions using design tools, to be completed entirely on the generation AI, contributing to faster content creation and reduced workload.
SOURCE: PRTimes


