AI & MCP
The missing layer between AI and your enterprise data.
Every company is told to do AI — but assistants without data access produce confident fiction, and assistants with raw database credentials are a breach in waiting. Virtual Data Platform's MCP server is the governed middle.
- Live, governed data for AI
- Per-user permissions
- Fully audited
How it works
Assistants discover Virtual Datasets as tools and query live — per-user identity, auth translation, sandboxing and full auditability, identical to human access.
MCP → Virtual Dataset → Agent → source
Discovered as tools
Assistants see the Virtual Datasets you license as callable tools — nothing else is exposed.
Run under the user's identity
Each query inherits the signed-in user's permissions and auth translation — never a service account.
Sandboxed and audited
Queries run live through an ephemeral per-user sandbox; every one is attributable, like human access.
What this unlocks
Conversational controlling
Ask your numbers in plain language and get answers from live data.
Agentic workflows
Agents gather actuals across ERP systems and draft analyses, with governance intact.
AI features, no data layer
Point any MCP-capable app at the datasets you expose — identity and audit included.
Variance commentary eats the first week of every month.
With Virtual Data Platform, an AI agent pulls actuals against budget across systems through the MCP server and drafts the deviation analysis; the controller reviews and signs off, every query attributable to a user.
The agent does the gathering, the human does the judging.
- MCP
- AI agent
- Audit
An example conversation
Open order backlog for plant Graz, by week?
vdp.query("orders_backlog", plant="Graz")Here is the weekly open order backlog for plant Graz — 1,284 open orders across 9 weeks, peaking in week 24.
Executed live against SAP, with the asker's own permissions.
Works with anything that speaks MCP
Tested in Azure AI Foundry. Claude, Microsoft Copilot Studio and custom agents are all examples of MCP clients — connect any of them.
Frequently asked
No. The assistant inherits the signed-in user's identity; source-system permissions apply unchanged.
See your own data live in 30 minutes.
Start with a live demo — we'll show your use case in Excel, Power BI, R, or through an AI assistant. Then take it further with a free trial in your own isolated tenant.