01 / KNOWLEDGE
Curated knowledge base
Admins build and manage the knowledge employees can query — documents, drives, and internal sources, organised by team.
Musashino RAG · Enterprise AI workspace
AI chat and knowledge,governed by youradmin console.
MRAG lets teams chat with files, web, and internal knowledge. Admins govern access, sources, use cases, logs, and audit trails from a single console.
The problem
AI is powerful.But enterprise AI is fragmented.
- 01Answers without accuracy checks.
- 02Sources without boundaries.
- 03Usage without oversight.
Answer layerNo accuracy check
Source layerNo boundaries
Control layerNo audit
Three gaps. One governed platform.
WHAT MRAG IS
Not a chatbot.
Not a RAG library.
A managed AI workspace for enterprise teams.
MRAG brings daily AI work and enterprise control into one platform. Teams chat with files, web, and internal knowledge. Admins manage access, approved use cases, knowledge sources, and what AI is allowed to do — from a single console.
02 / CONTROL
Admin-defined boundaries
Each use case has its own scope: which sources, which teams, what the AI is allowed to do.
03 / VERIFY
Answers you can trust
Every answer is grounded in your knowledge base and traceable to its source.
04 / OVERSIGHT
Full visibility
Admins can read every conversation, see what sources were used, and tune or revoke access at any time.
Core capabilities
Six capabilities, one platform.
01
AI chat with your knowledge
Employees ask questions and get answers drawn from your organisation's knowledge base — not the open internet.
04
Multi-step AI agents
For complex questions, MRAG breaks the task into steps, retrieves from the right sources, and assembles a verified answer — automatically.
05
Citation-linked knowledge
Index files, Google Drive, and internal documents. Every answer links back to the document and page it came from.
02
Session-private file uploads
Users can attach files during chat. Those files stay private to the session, separate from the shared knowledge base.
03
Web search, admin-controlled
Web access can be enabled or disabled per use case by the admin.
06
Admin-controlled governance
Use cases, access groups, retention policies, model selection, and full audit logs. The control plane that makes the other five deployable.
01
AI chat with your knowledge
Employees ask questions and get answers drawn from your organisation's knowledge base — not the open internet.
04
Multi-step AI agents
For complex questions, MRAG breaks the task into steps, retrieves from the right sources, and assembles a verified answer — automatically.
05
Citation-linked knowledge
Index files, Google Drive, and internal documents. Every answer links back to the document and page it came from.
02
Session-private file uploads
Users can attach files during chat. Those files stay private to the session, separate from the shared knowledge base.
03
Web search, admin-controlled
Web access can be enabled or disabled per use case by the admin.
06
Admin-controlled governance
Use cases, access groups, retention policies, model selection, and full audit logs. The control plane that makes the other five deployable.
Two connected experiences
Where your team works. Where your admin decides.
MRAG is not one product with two skins. It is two purpose-built surfaces that share the same knowledge engine, so what your team uses and what your admin controls are always in lockstep.
01 / User AI Workspace
For the people who do the work
Chat with your knowledge, files, and the web. Switch use cases, see source citations on every answer, recall full chat history. The same interface from desktop to mobile, English and Japanese, light and dark.
- Pick a use case — MRAG applies the right knowledge and scope automatically
- Files stay session-private; shared knowledge stays organisation-wide
- Every answer shows the source it came from
- Full chat history, searchable and recallable
One knowledge engine. Two surfaces. Always in lockstep.
02 / Admin Console
For the people who are accountable
Build knowledge bases, configure use cases, manage access groups, monitor every conversation. Define what AI can do, who can use it, and what it can draw from — without touching the chat surface.
- Set which use cases each team can access, and what the AI is allowed to do inside each one.
- Connect and index knowledge sources, then control which groups can query them.
- Read every conversation and see which sources were used — without touching the chat surface.
- Configure SSO, MFA, retention windows, and audit exports to satisfy your compliance team.
Two connected experiences
Where your team works. Where your admin decides.
MRAG is not one product with two skins. It is two purpose-built surfaces that share the same knowledge engine, so what your team uses and what your admin controls are always in lockstep.
01 / User AI Workspace
For the people who do the work
Chat with your knowledge, files, and the web. Switch use cases, see source citations on every answer, recall full chat history. The same interface from desktop to mobile, English and Japanese, light and dark.
- Pick a use case — MRAG applies the right knowledge and scope automatically
- Files stay session-private; shared knowledge stays organisation-wide
- Every answer shows the source it came from
- Full chat history, searchable and recallable
One knowledge engine. Two surfaces. Always in lockstep.
02 / Admin Console
For the people who are accountable
Build knowledge bases, configure use cases, manage access groups, monitor every conversation. Define what AI can do, who can use it, and what it can draw from — without touching the chat surface.
- Set which use cases each team can access, and what the AI is allowed to do inside each one.
- Connect and index knowledge sources, then control which groups can query them.
- Read every conversation and see which sources were used — without touching the chat surface.
- Configure SSO, MFA, retention windows, and audit exports to satisfy your compliance team.
Why MRAG is different
- Not a chatbot.
- Not a RAG library.
- Not a clever model.
- Not generic AI.
- Not one rigid flow.
A managed AI workspace,built on your organisation's knowledge,configured by admins,used by teams.
Answer states: grounded · needs improvement · low confidence ·missing source · clarify · fallback ·
How MRAG works
Configure once, govern continuously.
Admin shapes. Team uses. Admin improves.
ADMIN SHAPES
TEAM USES
ADMIN IMPROVES
- 01
Users & access
SSORoles - 02
Knowledge
IndexPartition - 03
Use cases
PromptsGuardrails - 04
Capabilities
FilesWebAgents - 05
Teams chat
CitationsVerdicts - 06
Audit & tune
TraceRevoke
CONTINUOUS GOVERNANCE LOOP
Observe, learn, and improve — then repeat.
ADMIN SHAPES + ADMIN IMPROVES
- 01
Users & access
SSO, MFA, roles, retention.
- 02
Knowledge
Index, tag, partition by group.
- 03
Use cases
Define prompts, scope, and guardrails per team.
- 04
Capabilities
Enable file search, web access, and agent tools.
- 06
Audit & tune
Read every trace. Tune. Revoke.
TEAM USES
05
Teams chat
Answers with sources, inline.
CONTINUOUS GOVERNANCE LOOP
Questions
FAQ
Those are general-purpose AI assistants — you bring your own context every time. MRAG is a managed platform where admins build curated knowledge bases from your organisation's documents and sources. Employees then get AI answers drawn from that knowledge, not from the open internet. Everything is scoped, logged, and reviewable.
Yes. The admin console gives you full chat history across all users — who asked what, which sources were used, and what the AI answered. You can review any conversation at any time without touching the chat surface.
MRAG runs as a managed SaaS with isolated workspaces per organisation. Your data is not shared across customers. For companies with stricter requirements, dedicated single-tenant deployments are available on request.
Initial setup is handled with you — building knowledge bases, configuring use cases and prompts, connecting data sources, and onboarding users. After that, your admin manages everything from the console: updating knowledge, adding users, adjusting access per team. No engineering required for day-to-day operation.
Yes. Each use case is configured separately — different prompts, different data sources, different access groups. A hotel's front desk, a bank's compliance team, and a sales team at a software company can all run on the same platform with completely separate configurations.
Start with a 30-minute walkthrough on your own data. If it fits, we run a paid pilot: one team, two use cases, four weeks. Pilots convert into full deployments without re-procuring.
Ready when you are
AI your team trusts. Knowledge your admin controls.
Book a 30-minute walkthrough. We show you how MRAG works on your own data, end to end.
MRAG · Musashino