AI agent governance, built in Japan
AI agents are already reading your systems and acting on your data. RAYVEN builds the identity and permission layer that lets regulated enterprises adopt them without losing control — so security teams never have to switch AI off to stay safe.
Two lines of business, one thesis: an AI agent is only useful in an enterprise once you can say what it is allowed to do.
We build and operate our own software. Tumiki is our core platform for governing what AI agents are allowed to do.
We embed our engineers with the customer to design, build and operate agent workflows end to end — a forward-deployed engineering model rather than a spec-and-handover contract.
Software we build, ship and operate ourselves.
Core platform
Tumiki gives AI agents an identity and a scope of authority, the way an organisation gives an employee a badge. Every connection an agent makes to an internal system goes through a permission boundary that an administrator defined and an auditor can inspect.
An agent can only take over work that somebody has described, and most organisations have never written theirs down. WorkLens AI measures how the work is actually done — from observed usage rather than self-reported timesheets — and turns it into the operating context an agent is built on. The same measurement then shows what the AI you deployed is returning, in hours and in yen, and keeps showing it month after month. Results are exposed over MCP, so an agent can read them directly.
Most AI agent projects do not fail on the model. They fail on the boundary — nobody can decide what the agent may touch, so the pilot never leaves the sandbox. We work the other way round: we start from the permission boundary and build outwards, with our engineers sitting inside the customer's team rather than behind a specification.
A RAYVEN engineer works alongside your own people for the length of the engagement — in your codebase, your review process and your incident channel. We are accountable for the workflow running in production, not for a document describing it.
We sit with the team that does the job today and write down what an agent would actually have to read, call and change to do it. That list, not a model choice, is the real scope.
Every item on that list becomes an explicit permission: who granted it, what it covers, and what evidence it leaves behind. This is where Tumiki does the work, and where a security review stops being a negotiation.
We implement the workflow on top of Tumiki or on your existing stack, ship it into production behind the boundary we designed, and stay on to operate it while your team takes it over.
The goal is not a dependency. We leave behind a workflow your engineers own, a permission model your auditors can read, and the reasoning behind both.
Granted for our management foundation technology around the Model Context Protocol (MCP). A PCT international application has been filed.
MCP is how agents reach the tools and data they act on, which makes it the right place to enforce permissions. Our GENIAC-PRIZE submission — "integrated risk countermeasures through MCP server hardening and AI permission control" — was selected as one of eight from sixty-seven entries.
RAYVEN operates MCP Japan, the largest MCP community in Japan with more than 15,000 members, and publishes engineering write-ups through its Qiita organisation.
What a buyer's security team needs to know before Tumiki touches a system of record. How your data is stored and used is agreed contractually — ask us and we will walk you through the terms.
Recognition from public agencies, financial institutions and industry programmes in Japan, largely for our work on generative AI safety and governance.
March 2026 · Kohnan Foundation
Grand Prize, 7th Business Model Award 2026
第7回ビジネスモデル・コンテスト Business Model Award 2026 最優秀賞
March 2026 · METI / JETRO
Grand Prize, SHIDO Next Innovator Demo Day
始動 Next Innovator デモデー 最優秀賞
March 2026 · Senshu Ikeda Bank
Excellence Award (ICT), 26th Senshu Ikeda Bank New Business Grant
池田泉州銀行 第26回ニュービジネス助成金 優秀賞（ICT部門）
February 2026 · The Kiyo Bank
Technical Award, 11th Kiyo Innovation Support Program
第11回 紀陽イノベーションサポートプログラム 部門賞（テクニカル賞）
December 2025 · Osaka Innovation Hub / NICT
NICT Award and Kinki Bureau of Telecommunications Director-General's Award, Mirai no Pitch 2025
ミライノピッチ2025（一般の部）NICT賞 / 近畿総合通信局長賞
November 2025 · Headline Japan (IVS)
Finalist, LAUNCHPAD SEED 2025
LAUNCHPAD SEED 2025 決勝進出（Finalist）
October 2025 · NEDO / METI
Award, GENIAC-PRIZE AI safety track (trial review)
GENIAC-PRIZE（安全性領域）トライアル審査 受賞
August 2025 · All-Osaka Entrepreneur Support Project
Nomura Securities Special Sponsor Award and SoftBank DX Supporter Award, 22nd Dream DASH! Business Plan Contest
第22回ビジネスプランコンテスト「ドリームDASH！」野村證券特別協賛賞・ソフトバンクDXサポーター賞（2賞同時受賞）
RAYVEN Inc. is a Japanese software company headquartered in Osaka. We build the permission layer that enterprises need before they can put AI agents anywhere near a system of record.
Full company profileWhether you are evaluating Tumiki, scoping an agent workflow, or looking for an engineering partner in Japan, we reply to enquiries within two business days.
