What MCP is, in plain words
Many people already work with an AI agent every day — Claude, Microsoft Copilot or another. These agents are good at writing, summarising and thinking a problem through. What they do not know is your business: this quarter's sales, what a customer owes, how much stock is left. That lives in your ERP and your other systems.
MCP (Model Context Protocol) is an open standard that lets an AI agent connect to outside tools and data in a consistent way. Think of it as a standard plug: a system that speaks MCP can be plugged into any agent that supports it, instead of a separate integration for each agent.
WizBI speaks MCP. Once it is connected, the AI agent your people already use can ask your company brain a question and get the answer from your own data.
What it looks like
A finance manager is preparing for a board meeting in an AI agent. She asks for the numbers; the agent asks WizBI; WizBI answers from Priority with her permissions; the agent writes the text around them.
The agent does the writing. WizBI supplies the numbers — from the source, and only what this person is allowed to see.
Why through WizBI, and not straight to the database
You can connect an AI agent directly to a database. For a developer exploring one database, a generic database connector may be all they need.
For a company it is usually not enough:
- Tables are not answers. An ERP has a great many tables and cryptic field names. WizBI knows your business terms — what an "open order" or "gross margin" means in your company.
- Permissions per person. WizBI applies the permissions of the person who is asking, the same ones they have in WhatsApp or Teams.
- More than one system. One question can draw on the ERP, monday.com and a spreadsheet together.
- Controlled and logged. The agent gets only the tools agreed with you, under your permissions, and every question is logged and can be reviewed.
For IT: how the connection works
WizBI acts as the MCP server in front of your company brain; the AI agent is the client. When someone asks the agent a business question, the agent calls WizBI. WizBI queries the connected sources, applies that person's permissions, logs the question and returns the answer. The agent does not get direct access to your ERP or databases.
- Sources: Priority ERP (fully supported), SAP, monday.com, Google Sheets, Excel and CSV files, BigQuery, Morning, SharePoint, Dropbox, SQL databases, documents and APIs. Cloud systems use a dedicated user or an API token; on-premise databases connect through a secure agent, with no inbound ports.
- Identity and permissions: set by organisation, group and user. The same person has the same permissions on every channel — WhatsApp, Microsoft Teams, Slack, Telegram, e-mail, the web and AI agents via MCP.
- Tools: the MCP tools WizBI exposes are the ones agreed with you, defined around your process and your permissions.
- Audit: every question is logged and can be reviewed.
- Platform: WIZBI AI LTD is certified to ISO/IEC 27001:2022. WizBI runs on Google Cloud; each organisation's data is isolated and encrypted in transit and at rest. The AI models WizBI uses run inside Google Cloud under enterprise terms, and your data is not used to train them.
One point to decide on your side: once WizBI returns an answer to an outside AI agent, that answer is handled by the agent under your organisation's agreement with its provider, so check that provider’s current terms. Choose which agents to connect the way you approve any other tool.
MCP or chat — which channel for whom
MCP does not replace the chat channels. Most of a team will keep asking in WhatsApp or Microsoft Teams, where they already are. MCP is for the people who already spend their day in an AI agent — finance, analysts, managers who write reports and presentations — and want the company's numbers in the same window.
Working with Priority? See AI for Priority ERP. Looking at public AI chat tools for company data in general? See ChatGPT and company data and AI agents for business.