How to use ChatGPT with MRPeasy
Short answer
You can use ChatGPT with MRPeasy by uploading exports for one-off analysis, or by building a custom GPT with an action that calls the MRPeasy REST API for live answers. For regulated or customer-confidential data, a private model behind the same API is the safer pattern.
Most MRPeasy users first try ChatGPT by exporting a stock or order list to Excel and dragging it into a chat. That works for a quick question, but it goes stale the minute someone ships an order, and it puts a snapshot of your customers and prices in a third-party system.
The next step is live access. ChatGPT supports custom GPTs with actions defined by an OpenAPI schema, which means it can call a service you control. That service, not ChatGPT, talks to MRPeasy, and decides what the model is allowed to see.
Whether this is appropriate depends on your data and your plan. Consumer plans differ from business and enterprise plans in training and retention terms, so read the current terms before sending anything sensitive.
What teams use it for
Ad hoc stock analysis
Upload a stock export and ask which items have more than a year of cover or have not moved, as a starting point for a cycle count or write-off discussion.
Explaining a late order
Ask a custom GPT why customer order 1234 is late. It calls your action, fetches the order and manufacturing orders, and explains the chain.
Drafting customer updates
Paste an order status and ask for a polite delay notice with the new date, in the customer's language.
BOM cost comparison
Export two BOM versions and have ChatGPT summarise what changed and the cost effect for an engineering change review.
Spreadsheet cleanup
Use it to reshape a supplier price list into the import format MRPeasy expects, then validate before importing.
Report formulas
Ask for Excel or Power Query steps to turn an MRPeasy export into an on-time delivery summary.
MRPeasy integration surface
- CSV and Excel exports from lists and reports in the MRPeasy UI. Fastest path, no engineering, but manual and point-in-time.
- REST API with API key and secret authentication, usable as the backend of a custom GPT action. Exact endpoints and field names vary by version, so build your OpenAPI file from your account's current API documentation.
- MRPeasy import tools for items, BOMs and prices, useful when ChatGPT cleans data that you then load back through the supported import.
- No direct database access because the product is hosted. Any live integration is API-based.
- Rate limits on the API, which matter if a custom GPT triggers many calls from a single question.
How it works
- 1
Decide the data boundary
List the fields that can leave your environment, such as item codes, quantities and dates, and the ones that cannot, such as customer prices or contact details.
- 2
Start with exports
For analysis tasks, export a filtered list, remove sensitive columns, and upload it. Treat the result as advice to verify.
- 3
Build a small middle service
For live access, host an HTTPS service that exposes a few read-only endpoints, for example order status and stock by item, and calls MRPeasy server-side.
- 4
Describe it with OpenAPI
Write a minimal schema for those endpoints and add it as an action to a custom GPT, with authentication to your service rather than to MRPeasy.
- 5
Write tight instructions
Tell the GPT to cite the order numbers it used, to say when data is missing, and never to guess dates.
- 6
Test against known cases
Compare answers with what the planner sees in MRPeasy for 20 real orders before sharing the GPT.
Example flows
Why is this order late
Trigger: A salesperson asks the custom GPT about a customer order number.
- The GPT calls the order-status action with the order number.
- Your service fetches the customer order, linked manufacturing orders and open purchase orders from MRPeasy.
- The service returns a compact summary of delays and missing components.
- The GPT explains the cause and the next dated event.
Result: The salesperson gets a plain answer in seconds without asking the planner.
Slow-mover review
Trigger: The purchasing lead uploads a stock export before a quarterly review.
- Remove cost columns if policy requires.
- Ask ChatGPT to group items by movement and months of cover.
- Review the proposed list and check a sample in MRPeasy.
- Export the agreed list for the cycle count team.
Result: A review prep job that took an afternoon takes under an hour.
Keeping data on-prem
MRPeasy is hosted, so what stays on-prem is the middle service and, if required, the model. ChatGPT is a cloud product, so any prompt and returned data travel to OpenAI. If your contracts or export rules forbid that, keep the same architecture but swap the model: run an open-weight model on your own server behind a chat front end, and have it call the same read-only service. You keep the conversational experience and the MRPeasy API boundary, and the data never leaves infrastructure you control. Hybrid is also reasonable: ChatGPT for non-sensitive spreadsheet help, a private model for anything touching customers or defense work.
Security checklist
- Never put the MRPeasy API secret inside the custom GPT. Authenticate the GPT to your own service only.
- Expose read-only endpoints and filter sensitive fields in your service, not in the prompt.
- Check the plan's data usage and retention settings before uploading any export.
- Add per-user authentication to the service so the GPT cannot become a shared back door.
- Keep an access log of every action call with the user and the order or item queried.
Ways to get it done
Manual export and upload
Effort: Minutes
Occasional analysis with non-sensitive data.
Custom GPT with a self-hosted action service
Effort: 1-3 weeks
Teams on ChatGPT Team or Enterprise who want live answers to simple questions.
Private model with Netray
Effort: Typically a few weeks for a scoped pilot
Regulated, customer-confidential or export-controlled data, or where you want approvals and audit.
Pitfalls
- !Uploading full customer lists with prices to a personal account.
- !Trusting calculated answers from a large export. Models miscount, so spot check totals.
- !Hard-coding the API secret into the GPT configuration.
- !Letting one GPT expose every endpoint. Build per-role GPTs or enforce roles in the service.
- !Forgetting that an export is stale. Label the extraction time in the file name and in the prompt.
Frequently asked questions
Is there an official ChatGPT connector for MRPeasy?
At the time of writing there is no widely known official one, but listings change. Check the MRPeasy integrations page and ChatGPT's app directory. If none exists, a custom GPT action over the REST API or an MCP-based connector is the practical route.
Can ChatGPT write directly into MRPeasy?
It can if you build an action that does so, but we advise against it at first. Keep actions read-only, and when you add a write such as creating a draft record, require a human confirmation step in your service.
Will OpenAI train on my MRPeasy data?
Policies depend on the plan and settings and change over time. Business and enterprise offerings generally state that inputs are not used for training by default, but read the current terms and your contract rather than assuming.
Is a custom GPT or the API better?
A custom GPT is quickest for a small internal team. If you want it embedded in your own portal, with your own login and logging, call the model through the API from your own application instead.
What if our customer contract bans cloud AI?
Then use a private model on your infrastructure. The MRPeasy API side stays the same, only the model moves on-prem, so the work you did on the integration layer is reusable.
Want ChatGPT working on your MRPeasy?
A free 30 minute assessment: we map the integration path for your version and setup, and where it should run.
Get a free assessment