How to Connect PostgreSQL to Microsoft Copilot with Windsor.ai

To connect PostgreSQL to Microsoft Copilot, connect your database in Windsor.ai, add Windsor MCP as a Model Context Protocol tool in Microsoft Copilot Studio, then add the tool to your agent and ask about your data in plain language. Setup takes about a minute, needs no code, and stays read-only, so Copilot can read and report on your tables without ever changing anything in your database.

Once connected, anyone on your team can ask Copilot questions about your live PostgreSQL data and get answers in seconds, with no SQL, no manual exports, and no waiting on an analyst.

• No code and no SQL. You ask in plain language and Copilot does the querying for you.
• Live data on every request, so answers reflect what is in your tables right now, not a stale export.
• Read-only by design. Windsor.ai reads and reports your data and never writes back or changes anything at the source.
• Self-serve for the whole team, without depending on a data analyst.
• Blend PostgreSQL with your ad, analytics, and CRM data for cross-channel reporting from one place.
• Free forever plan, no credit card. Paid plans start from 19 dollars per month.

Before you start

You will need three things:

• A Windsor.ai account. The free forever plan is enough to get started, and no credit card is needed.
• Access to your PostgreSQL database, including the host, port, database name, username, and password, so Windsor.ai can read your tables.
• Access to Microsoft Copilot Studio, where you build and manage your Copilot agents.

Connect PostgreSQL to Microsoft Copilot with Windsor.ai

The setup has four short steps.

Step 1: Connect PostgreSQL in Windsor.ai

Go to onboard.windsor.ai/app/postgresql, sign in, and authorize PostgreSQL by entering your read-only connection details (host, port, database, user, and password). Windsor.ai connects to your database and reads only the tables you grant access to. Nothing is written back.

Step 2: Add Windsor MCP as a tool in Copilot Studio

In Microsoft Copilot Studio, open or create an agent, then go to Tools. Click Add a tool, then New tool, then Model Context Protocol. Fill in the details as follows:

• Server name: Windsor MCP
• Server URL: https://mcp.windsor.ai
• Authentication: API Key
• Header name: Authorization
• Header value: Bearer YOUR_WINDSOR_API_KEY (replace this with your own Windsor.ai API key, keeping the Bearer prefix)

Click Save. You can find your API key inside your Windsor.ai account.

Step 3: Connect and authorize

Open the connections manager for the tool. Under the Windsor MCP connection, click Connect, then Submit. If the status does not update straight away, click Retry until it shows as connected.

Step 4: Add the tool to your agent and ask

Add the Windsor MCP tool to your agent, open a chat, and ask about your data in plain language. Copilot uses Windsor MCP to read your PostgreSQL tables and answer.

Ready to try it? Start for free with Windsor.ai. There is a free forever plan and no credit card, so you can connect PostgreSQL and start asking Copilot questions in minutes.

How Windsor.ai reads your PostgreSQL tables

Windsor.ai connects to your PostgreSQL database with the read-only credentials you provide and exposes your own tables and columns to Copilot through the Windsor MCP server. Because every database schema is different, there is no fixed list of fields. Instead, Copilot works with the tables and columns that already exist in your database, whatever you have named them. When you ask a question, Copilot turns it into a read query, Windsor.ai runs it against your live data, and the result comes back into the chat. Windsor.ai never writes, updates, or deletes anything in your database.

The PostgreSQL data that matters

Because this is your own database, the data that matters is whatever lives in your tables. Most teams point Copilot at a handful of core tables. Here are common examples and the kinds of questions they unlock.

Table (example)	What you can ask Copilot
Orders and revenue tables	Revenue by month, top products, average order value, refunds
Customers and accounts tables	New signups, active accounts, churn, lifetime value
Subscriptions and billing tables	Recurring revenue, upgrades and downgrades, outstanding invoices
Products and inventory tables	Best sellers, stock levels, category performance
Events and usage tables	Feature adoption, daily active users, usage trends
Support and ticket tables	Open tickets, resolution time, most common issues

Point Copilot at the tables and columns you already have, and ask about them by name.

What to ask Copilot: prompt ideas

Use the real table and column names from your database. A few examples to start with:

Show me total revenue by month for the last 12 months from my orders table, and tell me which month was strongest.

Summarize new customer signups per week from my customers table for the last 90 days, and flag any week with an unusual drop.

Which product categories had the highest sales last quarter in my orders table? Give me a ranked list with totals.

Can Microsoft Copilot connect directly to a PostgreSQL database?

Not on its own. Microsoft Copilot does not include a built in way to open a live conversation with an arbitrary PostgreSQL database. To bridge the two you need a connector that speaks a protocol Copilot understands. Windsor.ai does this through the Model Context Protocol (MCP): you connect PostgreSQL once in Windsor.ai, add the Windsor MCP tool in Copilot Studio, and Copilot can then read your tables in plain language. There are other routes for different Copilot products, which the next section compares.

Ways to connect PostgreSQL to Microsoft Copilot

The name “Copilot” covers several different Microsoft products, and each has its own way to reach a PostgreSQL database. Here is how they compare.

Method	Best for	Code needed	Notes
Windsor MCP in Copilot Studio (recommended)	Business teams asking questions in plain language	None	No-code, read-only, live data, connects in about a minute
Microsoft 365 Copilot PostgreSQL connector (Graph connector)	Indexing database content into Microsoft Search	SQL required	You write SQL queries and access control lists to index columns, so it needs more setup and is geared to enterprise search
GitHub Copilot in VS Code (@pgsql)	Developers writing and debugging SQL	Developer tooling	Schema-aware chat inside the editor, not a business reporting tool
Direct API and custom code	Engineering teams building their own integration	Full coding	Build and host your own connector or MCP server against the PostgreSQL API

For most marketing, revenue, and operations teams, the Windsor MCP route is the fastest and needs no code.

Is it safe? Permissions and read-only access

Yes. Two things keep your database safe:

• Read-only access. Windsor.ai only reads and reports your data. It never writes, updates, or deletes anything in your PostgreSQL database, so your tables cannot be changed through Copilot. For an extra layer of safety you can give Windsor.ai a database user that has read-only permissions.
• No credentials exposed to Copilot. Your PostgreSQL credentials stay inside Windsor.ai. Copilot connects to Windsor MCP with an API key, not to your database directly, so your database login is never shared with Copilot.

You control exactly which tables Windsor.ai can see, and you can revoke access at any time.

Troubleshooting the PostgreSQL to Copilot connection

• The tool shows as “not connected.” Open the connections manager, click Connect, then Submit, and click Retry until the status updates. The first handshake sometimes needs a second try.
• Copilot does not use the tool. Make sure the Windsor MCP tool is added to the specific agent you are chatting with, not only created.
• Authentication fails. Check that the header name is exactly Authorization and the value is Bearer followed by a space and your Windsor.ai API key.
• Copilot cannot find a table. Confirm the table exists in the database you connected and that your Windsor.ai connection has access to it, then refer to it by its exact name.
• No data comes back. Widen the date range or double check the column names, since Copilot queries the columns you actually have.

Bonus: analyze PostgreSQL alongside your full stack

Your database rarely tells the whole story on its own. Because Windsor.ai connects more than 350+ sources, you can blend PostgreSQL with your ad, analytics, and CRM data and ask Copilot cross-channel questions from one place:

• Blended ROAS: pair revenue from your PostgreSQL orders table with ad spend from Google Ads and Meta Ads to see true return on ad spend.
• Single source of truth: compare signups in your database with sessions in GA4 to reconcile the numbers your teams report.
• Cross-channel reporting: line up subscription revenue from your database against email and campaign performance to see what really drives growth.

This turns Copilot into a self-serve reporting layer over every source your team cares about, not just your database. Windsor.ai is used by 9,000+ marketing teams for exactly this kind of cross-channel reporting.

Connect PostgreSQL to other AI tools

Windsor MCP is not limited to Copilot. With the same read-only Windsor.ai connection you can also connect PostgreSQL to Claude or connect PostgreSQL to ChatGPT, so your team can ask about the same data in whichever assistant they prefer. To see everything Windsor.ai brings into Microsoft’s assistant, browse the Microsoft Copilot Agent integrations, and for a reference of the connector itself, see the PostgreSQL connector fields.

Conclusion

Connecting PostgreSQL to Microsoft Copilot with Windsor.ai gives your team plain-language access to live database data in about a minute, with no code and full read-only safety. Connect once, add the Windsor MCP tool to your Copilot agent, and start asking.

Connect your PostgreSQL database to Microsoft Copilot today. Start for free with Windsor.ai, with a free forever plan and no credit card, and ask Copilot your first question in minutes.