AI for a "Workable Environment" in Companies. -- MCP Zero Trust Gateway "Tumiki MCP" Pre-launches Function to Build Business-Specific Agents
RAYVEN Inc. today, April 30th, announced the pre-launch of a new feature for its MCP Zero Trust Gateway "Tumiki MCP," which automatically constructs business-specific virtual MCPs from business descriptions. This also strengthens AI client identification logs and PII masking functions. The company, an award winner of "Shido Next Innovator" supported by METI and JETRO, realizes an "AI Environment" within companies for AI agents using its patented technology (Patent No. 7731114).
📋 Article Processing Timeline
- 📰 Published: May 1, 2026 at 08:58
- 🔍 Collected: May 1, 2026 at 00:31
- 🤖 AI Analyzed: May 1, 2026 at 03:24 (2h 52m after Collected)
RAYVEN Inc. (CEO: Yoshihiro Suzuyama) today, April 30th, announced the pre-launch of a new feature for its MCP Zero Trust Gateway "Tumiki MCP," which automatically constructs business-specific virtual MCPs from business descriptions. AI client identification logs and PII masking functions have also been simultaneously enhanced. The company, an award winner of "Shido Next Innovator" supported by the Ministry of Economy, Trade and Industry (METI) and JETRO, realizes an "AI Environment" within companies for AI agents using its patented technology (Patent No. 7731114).
▼ Tumiki MCP LP: https://lp.tumiki.cloud/
Tumiki Landing Page
Without an "Employee ID" for AI, AI cannot have a "Workable Environment"
When a human joins a company, an employee ID is issued. This manages permissions—which rooms can be entered, which systems can be accessed—and records actions. An employee ID is the minimum infrastructure for an organization to trust an individual and assign work.
However, currently, AI agents "joining" companies lack this mechanism. Because there is no employee ID, companies have not yet been able to build a safe working environment (AI Environment) for AI.
The era has begun where generative AI tools like Claude Code, Cursor, ChatGPT, and Gemini autonomously connect to internal systems via MCP to perform tasks. However, companies lack a mechanism to control "who can assign what tasks to which AI."
RAYVEN has been developing "Tumiki MCP," an MCP Zero Trust Gateway that issues employee IDs to AI and builds a secure execution environment where AI can work safely. The business-specific virtual MCP construction function, launched today, enables anyone in a company to build business-specific AI agents on top of this environment.
Why now? — The rapid spread of MCP accelerates corporate risks
MCP (Model Context Protocol), released by Anthropic and subsequently supported by major AI companies such as OpenAI, Google, and Microsoft, is becoming the de facto standard protocol for connecting AI agents and enterprise systems.
While the adoption of MCP is spreading to major API gateway products, MCP also carries security and governance risks similar to existing APIs. Vulnerabilities in MCP servers leading to Tool Poisoning, privilege escalation, and unintended transmission of confidential information—these are challenges too burdensome for individual companies to address.
MCP Tool Call Logs
At the same time, there is another challenge: the feedback from the field that "even with MCP tools, it's difficult to assemble them into business operations." There are individual tools like GitHub's MCP, Slack's MCP, and Salesforce's MCP, but combining them to create operations such as "subsidy application," "performance evaluation," or "sales opportunity review" requires significant engineering effort.
Tumiki MCP solves both governance and business-specific virtual MCP construction on a single platform.
Build business-specific virtual MCPs with just business descriptions
Tumiki MCP provides a function that automatically selects and optimizes necessary functions from a company's MCP toolset based on natural language descriptions of business content, thereby constructing business-specific virtual MCPs.
Business-specific virtual MCPs are completed in 4 steps
Step 1 | Describe business in natural language: Input the tasks to be performed, such as "I want to automate subsidy applications" or "I want to conduct objective performance evaluations."
Step 2 | Automatically select necessary tools: Automatically pick up tools necessary for the task within the user's MCP permissions. Tools outside of permissions will not be selected.
Step 3 | Optimize for business context: Automatically overwrite and optimize the description of each tool according to the input business context. This maximizes the accuracy of AI agent calls when performing tasks.
Step 4 | Immediately operate as a business-specific virtual MCP: The optimized toolset is bundled as a business-specific "virtual MCP" and becomes immediately available via Tumiki MCP.
Build Virtual MCP
"Three Control Layers" provided by Tumiki
Monitor | Real-time visualization of all communications
Who, when, with which AI, and what service was accessed. Everything is recorded in real-time, and unauthorized access is immediately blocked.
Control | Detailed permission settings for each role
Read / Write / Execute permissions are finely set for each position and department.
▼ Tumiki MCP LP: https://lp.tumiki.cloud/
Tumiki Landing Page
Without an "Employee ID" for AI, AI cannot have a "Workable Environment"
When a human joins a company, an employee ID is issued. This manages permissions—which rooms can be entered, which systems can be accessed—and records actions. An employee ID is the minimum infrastructure for an organization to trust an individual and assign work.
However, currently, AI agents "joining" companies lack this mechanism. Because there is no employee ID, companies have not yet been able to build a safe working environment (AI Environment) for AI.
The era has begun where generative AI tools like Claude Code, Cursor, ChatGPT, and Gemini autonomously connect to internal systems via MCP to perform tasks. However, companies lack a mechanism to control "who can assign what tasks to which AI."
RAYVEN has been developing "Tumiki MCP," an MCP Zero Trust Gateway that issues employee IDs to AI and builds a secure execution environment where AI can work safely. The business-specific virtual MCP construction function, launched today, enables anyone in a company to build business-specific AI agents on top of this environment.
Why now? — The rapid spread of MCP accelerates corporate risks
MCP (Model Context Protocol), released by Anthropic and subsequently supported by major AI companies such as OpenAI, Google, and Microsoft, is becoming the de facto standard protocol for connecting AI agents and enterprise systems.
While the adoption of MCP is spreading to major API gateway products, MCP also carries security and governance risks similar to existing APIs. Vulnerabilities in MCP servers leading to Tool Poisoning, privilege escalation, and unintended transmission of confidential information—these are challenges too burdensome for individual companies to address.
MCP Tool Call Logs
At the same time, there is another challenge: the feedback from the field that "even with MCP tools, it's difficult to assemble them into business operations." There are individual tools like GitHub's MCP, Slack's MCP, and Salesforce's MCP, but combining them to create operations such as "subsidy application," "performance evaluation," or "sales opportunity review" requires significant engineering effort.
Tumiki MCP solves both governance and business-specific virtual MCP construction on a single platform.
Build business-specific virtual MCPs with just business descriptions
Tumiki MCP provides a function that automatically selects and optimizes necessary functions from a company's MCP toolset based on natural language descriptions of business content, thereby constructing business-specific virtual MCPs.
Business-specific virtual MCPs are completed in 4 steps
Step 1 | Describe business in natural language: Input the tasks to be performed, such as "I want to automate subsidy applications" or "I want to conduct objective performance evaluations."
Step 2 | Automatically select necessary tools: Automatically pick up tools necessary for the task within the user's MCP permissions. Tools outside of permissions will not be selected.
Step 3 | Optimize for business context: Automatically overwrite and optimize the description of each tool according to the input business context. This maximizes the accuracy of AI agent calls when performing tasks.
Step 4 | Immediately operate as a business-specific virtual MCP: The optimized toolset is bundled as a business-specific "virtual MCP" and becomes immediately available via Tumiki MCP.
Build Virtual MCP
"Three Control Layers" provided by Tumiki
Monitor | Real-time visualization of all communications
Who, when, with which AI, and what service was accessed. Everything is recorded in real-time, and unauthorized access is immediately blocked.
Control | Detailed permission settings for each role
Read / Write / Execute permissions are finely set for each position and department.