Zoined MCP
Zoined MCP
Your Zoined data, in Claude and ChatGPT.
Zoined's MCP (Model Context Protocol) is a connector that lets an AI assistant ask Zoined for results. The assistant asks the question, Zoined does the calculation. No manual exports and no files pasted into a chat.

Example from our clothing demo data connected to ChatGPT.
What you can ask


How did sales develop last week compared to the same week last year?
Uses your company’s own time periods and comparison periods.


Which stores grew most this month compared to last year? And which declined?
Results come back ranked by change.


Margin % by product category for the last 3 months compared to budget.
Metrics by grouping, against budget.


Monthly sales this year in one store.
Finds the store by name and filters to it.
Follow-up questions build on the earlier answers.
How it works
1
Your data is in Zoined.
POS and other sources flow into Zoined’s data warehouse, as they do today.
2
Zoined holds the definitions.
What net sales means, whether returns are in it, and how each metric is calculated for one store and for the chain.
3
The assistant asks, Zoined calculates.
The assistant requests a result and explains what comes back.
Why not connect Claude directly?
Claude and ChatGPT can already read a sales export or query a database. What they don’t have is your business rules, so an answer can be wrong and still look right.
Claude on its own
Reads raw exports, or queries your POS database directly.
Guesses what net sales means and whether returns are in it.
Averages rates store by store, which gives the wrong chain figure.
Heavy questions can slow the systems your tills run on.
Sends thousands of raw records into the chat, which costs more.
Claude with Zoined
Asks Zoined, which already holds your cleaned, combined data.
Uses Zoined’s definitions for every metric, the same as your reports.
Gets chain totals calculated from the underlying numbers.
Queries run against Zoined’s warehouse, not your POS or ERP.
Receives finished results, so far less data goes into the chat.
Two stores convert 10% and 30% of their visitors. Averaged, that is 20%, but the chain’s real conversion is 25%, and an assistant working from a raw export won’t tell you which one it gave you.
Building your own MCP connects the assistant to the data. It doesn’t give it the definitions, the history or the warehouse behind them, which is the part Zoined already does.
Who sees what
Log in on Zoined’s own login page, with your password or SSO.
Each user gets only the data they can see in Zoined.
Admins choose which metrics and groupings AI tools can use.
Admins can turn access off for any user.
The assistant only reads data. It cannot change anything in Zoined.
What you need
A Zoined account. We switch the connector on for your company.
Claude on a paid plan, or ChatGPT on a plan with developer mode.
Not a Zoined customer yet? We can show you Zoined MCP on demo data in 20 minutes.
Limits
A very detailed question, such as a year of sales by product, returns a limited result. The assistant says so and suggests narrowing it.
For POS and system partners
If a customer forwarded you this page, here is what it means for your team.
Your customers can ask Claude or ChatGPT about their sales and get answers calculated by Zoined.
You build nothing. The connector is Zoined’s, and we switch it on for each customer.
Queries run against Zoined’s data warehouse, not your systems, so there is no extra load on your POS or API.
Data keeps flowing into Zoined through the integration you already have.
In pilot now
Zoined MCP is in pilot with customers. The first version is planned for Q4 2026.
