The ConverSight MCP server connects Claude, ChatGPT, and any MCP-compatible AI tool to your ERP data and the business context behind it. Ask about inventory, orders, suppliers, or margin, and get answers from your own numbers.
One governed connection between your business data and the AI tools your team chooses.
It uses MCP (Model Context Protocol), an open standard that lets AI assistants work with outside tools and data. Once it’s switched on, anyone on your team can ask Claude, ChatGPT, or another MCP-compatible tool about inventory, orders, suppliers, or margin. They get answers from your Digital Cube, and they only see what they already have access to.
One connection, any model. Your team picks the assistant. The data, definitions, and permissions stay the same.
ERP, warehouse, CRM, and planning data sit together in Digital Cube, so one question can draw on orders and inventory at the same time.
Datasets come with their relationships and calculations defined, so the assistant reads that structure before it writes a query.
The assistant queries your data in ConverSight directly. Nobody pulls a CSV first, and questions spanning years of history run in the background.
Guardrails
Plus the guardrails that decide who sees what
Your ConverSight account team turns on the MCP server for your organization.
The way you'd ask an analyst, in Claude, ChatGPT, or any MCP client.
Digital Cube tells the assistant where the answer lives in your data.
Every step shows in the chat, so anyone can see how a number was pulled.
Turn it into a chart, a leadership summary, or a what-if to test before it hits the plan.
Supply chain leaders
Planners & buyers
Sales
Finance
See the exposure before the Monday review.
Ask
AI assistant
ConverSight connected
Stop
Waiting on the weekly report to find out
See the exposure before the Monday review.
Which customers are behind on fill rate this quarter, and what’s driving it?
Waiting on the weekly report to find out
Where are we most exposed if our top supplier slips two weeks?
Stitching POs and on-hand stock together in Excel
Summarize last week’s backorders by region and chart the trend.
Building the weekly chart by hand
Get to the decision without building the spreadsheet first.
Which SKUs fall below reorder point in the next 30 days?
Checking SKUs one screen at a time
Show open POs past their promised date, grouped by supplier.
Scrolling the open PO report line by line
Compare actual lead times to the item master for our top 50 SKUs.
Guessing which lead times have gone stale
Answer the customer before you hang up.
What’s on backorder for my top five accounts, and when does it ship?
Emailing the warehouse for order status
Which of my top 20 customers ordered less this quarter than last, and on which products?
Pulling account history before every review
Which products are selling above forecast this month, and do we have stock to cover it?
Asking planning whether there’s stock to promise
Answer margin questions without filing a data request.
Margin by customer year to date versus last year. Who moved the most?
Rebuilding the margin sheet every month
What’s our inventory value by warehouse, and how much hasn’t moved in 180 days?
Filing a data request and waiting on it
Forecast versus actual by product family for the last six months.
Waiting on month-end close to see accuracy
ConverSight governance
More on how data stays governed: Trust & Security →
The assistant sees what that person can see in ConverSight. Dataset permissions and row-level security still apply, and nothing is shared between users.
Every tool starts on "Needs approval." Your team decides which read-only tools can run freely, and keeps anything that changes data behind a sign-off.
Any model can misread a question. Each tool it calls shows up in the conversation, so an analyst can check how a number was pulled before it reaches leadership.
Sign-in uses standard OAuth, tied to your organization's own domain. Disconnect at any time, and access ends immediately.
Connect Claude to your Digital Cube and ask from Claude's desktop or web app.
The same connection in ChatGPT, with the same data and permissions.
Other AI tools that support MCP connect the same way, through the same governed server.
Developers can point agents they build with MCP at your Digital Cube, under the same guardrails.
Prefer to keep everything inside ConverSight? Athena IQ reads the same Digital Cube, with no outside model involved.
The MCP server opens Digital Cube to the assistants your team already uses. Many teams start there, then add the rest as they grow.
Layer 3
Start from a ready-made blueprint, tailor it with a copilot, or describe a new app.
“ConverSight has bridged the gap between data silos, providing us with a unified decision intelligence platform that reduces decision-making time, enhances our agility in a competitive market, and drives efficiency across departments.”
“Before Athena, our on-premises data analytics lacked flexibility, required heavy maintenance and made data access difficult. Moving to the cloud seemed like a million-dollar, year-long project. With Athena AI, we delivered our first use case in under three months.”
“ConverSight’s AI insights enhanced the visibility and value of our ERP data, enabling faster and more informed decision-making.”
MCP (Model Context Protocol) is an open standard that lets AI assistants such as Claude connect to outside tools and data. ConverSight uses MCP to open the Digital Cube to Claude, so answers come from your business data, not the public internet.
No. Claude is one way in. Your team can also ask Athena IQ, ConverSight’s own AI assistant, inside every workspace and Microsoft Teams, or work in an app built in App Studio. All three read the same Digital Cube with the same permissions, so the answers match however people ask.
Only what the signed-in person can already see in ConverSight. Dataset permissions and row-level security apply to every question, and nothing is shared between users.
Not without approval. Every tool starts on “Needs approval,” so a person signs off before anything changes, such as merging a scenario branch.
Your ConverSight account team switches on the connector for your organization. After that, each person signs in with their work account and starts asking in Claude.
Yes. The same MCP connection works with ChatGPT and other AI tools that support MCP, with the same data and permissions.
Acumatica, Fishbowl, Sage, SAP, NetSuite, Dynamics, and Katana, among more than 150 connectors.
Answers come from the data in ConverSight, which refreshes as your systems sync. Repeat questions can be served from cache, so check the refresh time before a time-sensitive decision.
No. Ask the way you’d ask an analyst. The Digital Cube tells Claude where the answer lives, and Claude writes the query. Each step shows in the conversation, so an analyst can check how a number was pulled.
