DataSelf MCP+ Reference Guide

Intro

The DataSelf MCP+ for Data Warehousing & Analytics connector exposes your DataSelf data warehouse and analytics to any MCP-compatible AI assistant (such as Claude, ChatGPT, or other LLM-based tools) through a single, shared server URL secured with OAuth 2.0. Once connected, you sign in with your Microsoft account, and you can list tables, explore schemas, and query data using plain English — no SQL required.

In data warehousing (DW) environments, the Model Context Protocol (MCP) provides a standardized framework for AI systems to access, understand, and interact with warehouse data in a controlled and intelligent manner.

Key Concepts

• Entities — A connection may span multiple companies, divisions, or data warehouses. Each is an entity, identified by an Entity_Key. Most tools accept an optional entity_key to target a specific one; if omitted, the default connection is used.

• Schema-qualified tables — Tables and views are referenced as schema.table (for example dbo._F_Sales_Invoice). If the schema is omitted, dbo is assumed.

• Read-only by design — Only SELECT (or WITH … SELECT) statements are permitted. Anything that writes or alters data — INSERT, UPDATE, DELETE, MERGE, DDL, SELECT … INTO, EXEC, or multiple chained statements — is rejected before it reaches the warehouse.

• Row limits — Query results default to 1,000 rows and cap at 10,000. A TOP N clause is injected automatically when one isn’t specified, and very large responses (~500 KB+) are truncated.

SQL Server MCP — Setup Guide

This guide walks through connecting an MCP-compatible AI assistant to your SQL Server data warehouse.

Prerequisites

• A DataSelf Data Warehouse (Cloud, private-cloud, or on-premise) with MCP access enabled.

• A Microsoft account that DataSelf has granted access to your DataSelf Data Warehouse tenant.

• An MCP-compatible AI assistant or client — for example Claude, ChatGPT, or another LLM tool that supports custom MCP connectors.

This guide covers four steps:

• Step 1 — Confirm access and get the MCP Server URL

• Step 2 — Add a custom MCP connector in your AI assistant

• Step 3 — Sign in with your Microsoft account (OAuth 2.0)

• Step 4 — Start querying your data

Step 1 — Confirm Access and Get the MCP Server URL

The DataSelf MCP+ for Data Warehousing uses a single, generic server URL — the same URL is used by every customer:

https://mcp.dataself.com

Contact your DataSelf account manager or administrator to confirm that the Microsoft account you'll sign in with has been granted access to the correct data warehouse entities. Access is controlled entirely through your Microsoft account permissions.

Step 2 — Add a Custom MCP Connector in Your AI Assistant

Most MCP-compatible assistants provide a settings area for adding custom connectors or integrations — sometimes labeled “Connectors,” “Custom Integrations,” or “MCP Servers,” depending on the platform. Open that settings area in the assistant you use (for example Claude, ChatGPT, or another LLM tool) and choose the option to add a custom connector.

Fill in the connector form with the following values:

Confirm the connector to finish adding it.

Step 3 — Sign In With Your Microsoft Account (OAuth 2.0)

The connector authenticates using OAuth 2.0. The first time you use it, your AI assistant will open a Microsoft sign-in window. Sign in with the Microsoft account that DataSelf has granted access to, and approve the requested permissions.

DataSelf never sees or stores your Microsoft password — authentication is handled entirely by Microsoft's identity platform, and your assistant only receives a short-lived access token scoped to your account.

After adding the connector, your assistant will typically also ask how much tool access to grant. To allow full query capability, enable all actions — find the newly added DataSelf MCP+ for Data Warehousing connector in your integrations list, open its settings, and look for the Permissions or Actions section. Select “Always allow” (or the equivalent option that grants full tool access). This allows your assistant to list tables, run queries, and return results from your SQL Server.

Step 4 — Start Asking Questions

Open a new conversation. The DataSelf MCP+ for Data Warehousing connector is now active. Start with broad exploration questions and then go deeper from there.

Example prompts to get started:

• “Who am I connected as, and what entities do I have access to?”

• “What tables are available in the database?”

• “Show me total sales by month for the last 12 months.”

• “How many active customers do we have by region?”

• “What is the current inventory level by warehouse?”

• “Compare sales performance this year versus last year by product category.”

Your assistant may ask for your permission the first time it runs a query tool in a new session. Approve the request to proceed. Most assistants also let you expand a tool call to inspect the exact query that was executed.

Tool Reference

The connector exposes five tools. All of them are read-only.

whoami

Returns the authenticated user, the user type, and the current connection’s metadata. Use this first, in any new session, to confirm the login and see which DataSelf tenant the assistant is talking to.

list_entities

Lists every enabled entity the authenticated user can access — the companies, divisions, or warehouses in scope — ordered alphabetically by EntityID. Returns Entity_Key, EntityID, and Name_Entity for each. The Entity_Key returned here is used by the other tools to target a specific entity.

list_tables

Lists all tables and views in the Data Warehouse, formatted as schema.table. Scope it to a specific data warehouse with entity_key.

describe_table

Returns the column definitions for a given table or view, so the assistant can build accurate queries without guessing at a schema.

run_query

Executes a read-only T-SQL SELECT (or WITH … SELECT) against the warehouse. The statement is parsed before execution and rejected if it attempts to write data, run DDL, execute stored procedures, or chain multiple statements.

Example:

-- Top 5 customers by revenue this year

SELECT TOP 5 c.[Customer Name], SUM(f.SalesExt) AS Revenue

FROM dbo._F_Sales_Invoice f

JOIN dbo._D_Customer c ON f._CustomerFK = c._CustomerPK

WHERE f.[Invoice Date] >= '2026-01-01'

GROUP BY c.[Customer Name]

ORDER BY Revenue DESC;

Security & Access

• No write access — The connector can only run SELECT statements. Statements are parsed and rejected if they contain INSERT, UPDATE, DELETE, MERGE, DDL, SELECT … INTO, EXEC, or multiple chained statements.

• Permission-scoped — Access is limited to the entities and tables your authenticated DataSelf user is already entitled to see. The connector does not grant any additional privileges.

• OAuth 2.0 authentication — Every session requires signing in with a Microsoft account through the standard OAuth 2.0 consent flow. DataSelf never sees or stores your Microsoft password. Use whoami at any time to confirm which account and entity the assistant is operating under.

• Bounded results — Row counts and response size are capped to prevent unbounded data extraction in a single call.

For details on how DataSelf handles the data accessed through this connector, see the DataSelf MCP+ for Data Warehousing Privacy Policy.

Privacy Policy

Click here to learn more.

Limits

Support

Email: info@dataself.com

Phone: 888-910-9802

Knowledge Base: http://kb.dataself.com

Website: http://www.dataself.com