The MCP Server (Model Context Protocol) opens Xentral to modern AI applications. LLMs receive standardized and secure access to selected data and functions in your ERP system.
This allows you to connect tools such as ChatGPT, Claude Desktop, or Claude Code to Xentral and build AI-powered workflows directly on your business data.
Tipp
-
You can get started right away: Simply enter the MCP URL of your Xentral instance in ChatGPT or Claude (see Setup).
-
If you want to share feedback on business cases – "This would help me most in my business" – you are welcome to submit it here.
The Model Context Protocol (MCP) is an open standard for connecting AI systems to external applications and data sources.
Instead of exchanging information manually between different tools, LLMs communicate directly with Xentral and incorporate relevant information into their responses, analyses, or automations.
Depending on the permissions granted, LLMs can, for example:
-
Query ERP data: Ask about customers, orders, invoices, and stock levels in natural language – answers instantly, no menu navigation.
-
Uncover patterns and trends: Analyze revenue, return rates, or margin potential and dig deeper until you understand why.
-
Use reports: Find reports, run them, and export them as Excel or CSV files, without opening the report in Xentral. Running and exporting uses credits (see Using reports via the MCP Server).
-
Create and update records: Add customers, create orders, or update CRM entries directly – one sentence instead of a form. (not yet available)
-
Set up automations: Describe a process in plain language – Claude creates the workflows and agents directly in Xentral. (not yet available)
The available functions depend on the resources and permissions released for the MCP Server.
Examples
Instant answers in meetings
"How much revenue did we generate with customer Müller GmbH this month – and how did that compare to last month?" – Answer in seconds, straight from the live ERP, no login, no report to open. And if you want to dig deeper: "Show me just the last three months side by side" – no screen switch, no new search, just keep the conversation going.
Spot return patterns early
"Which products had an above-average number of returns this week?" – No pivot table, no pre-built report. And right away: "Which regions are they coming from?" – Claude goes deeper for as long as you keep asking. Ask this regularly and you catch patterns before they become problems.
Personalized dashboard as an artifact
"Show me a visual overview of my most important metrics for this week." – Claude pulls revenue, open orders, and returns directly from Xentral and renders an interactive dashboard in the chat. No setup, no configuration. And if you want more: "Add the top 3 customers by margin" – the dashboard grows with the question.
The MCP Server currently supports the following applications:
-
Claude Code
-
Claude Desktop
-
ChatGPT (Plus, Pro, or Business)
-
Codex
Support for additional MCP-compatible applications (e.g., Cursor, Windsurf) is planned for future versions.
Tip
First time using AI tools? No problem – just let Claude guide you through the setup. Open claude.ai and write something like:
"I want to set up the Xentral MCP Server. The MCP URL is: https://<your-instance>.xentral.biz/mcp – can you walk me through it step by step?"
Claude will take it from there. It may feel unfamiliar at first – that’s completely normal. The more you work with AI, the more natural it becomes.
Note
When connecting for the first time, a login dialog appears automatically. Authenticate once with your Xentral account – your connection is then saved for future use.
For all applications, use this MCP URL:
Replace <your-instance> with the name of your Xentral instance. This is the part before .xentral.biz in the address you use to open Xentral in your browser.
-
Open the settings in Claude Desktop.
-
Navigate to Connectors.
-
Click on Add custom connector.
-
Paste the MCP URL into the corresponding field.
In ChatGPT, you connect the MCP Server as a plugin. To do this, developer mode must be active.
-
In ChatGPT, open Settings → Security and login and activate Developer mode.
Whether you can activate developer mode depends on your ChatGPT plan and your workspace settings.
-
Open Plugins and click on the plus icon.
-
Enter a name, for example "Xentral", and a short description.
-
Paste the MCP URL under Connection.
-
Create the connection.
ChatGPT displays the tools that the MCP Server provides.
-
In Codex, open Settings and navigate to MCP servers.
-
Click on Add server.
-
Enter a name, for example "Xentral", select Streamable HTTP, and paste the MCP URL.
-
Save the server and restart Codex via Restart.
-
Click on Authenticate next to the Xentral server and sign in with your Xentral account.
Access is authenticated and based on the permissions defined in Xentral. LLMs are granted access exclusively to the data and functions that have been released for the respective connection.
This means you retain full control at all times over which information can be used by external AI applications.
Note
We recommend evaluating new use cases in a test environment first. Your feedback helps us continue to expand the integration and better support future AI workflows in Xentral.
Via the MCP Server, your AI application gets three kinds of tools:
-
Read-only ERP tools: access to the most important data in your Xentral, for example customers, sales orders, products, invoices, delivery notes, and returns, including analyses.
-
Tools for the Agent Hub: for example viewing your AI agents' tasks, managing agents, or maintaining your knowledge base.
-
Tools for reporting: for example running and exporting reports, creating your own reports, or scheduling recurring exports.
Your AI application displays which tools are available in detail after connecting. The range of functions is continuously expanded.
You can find, run, and export reports from reporting directly in a conversation with your AI application. This function is suitable if you want to retrieve key figures quickly without opening the report in Xentral, or if you want to have your own queries created. The tools are available in both the basic and the premium version of reporting.
Your access rights in reporting apply. If you have restricted access, you can only find, open, run, and export the reports of your access profile via your AI application. Xentral rejects other reports. Only users with full access can view credits, manage schedules and permalinks, run their own SQL queries, and create and change reports. You can find all details on user rights in the article Reports (new) - Data-driven reporting.
Note
You can only open reports from the [11] Real-time reports collection in Xentral, not via the MCP Server. Deleting reports, collections, and schedules is also only possible in Xentral.
Important
When your AI application runs a report or your own SQL query, or exports a report, this uses credits from your reporting, just as if you did it yourself in Xentral. The same rules apply: The same query with the same filter values uses only one credit per day, regardless of whether it is run or exported in Xentral or via your AI application.
A schedule uses one credit per report on every day it runs, regardless of how many recipients receive the export. Therefore, check how often the export should run before setting it up.
The preview, creating and changing reports, and creating schedules and permalinks are free of charge. If no credits are left, Xentral only rejects running reports and your own SQL queries as well as exporting, and does not charge anything. All other tools continue to work.
You can find all details on how credits are counted in the article FAQ Reports - Frequently Asked Questions.
On the first call that uses credits, Xentral pauses the process, does not use a credit yet, and asks your AI application to check with you. This also applies before a report is shared with new email addresses or via a link. Whether your AI application actually asks you depends on the application and your permission settings. A confirmation prompt is therefore not guaranteed.
Caution
A permalink works without signing in: Anyone who knows the link can retrieve the report data. The link is valid for a maximum of one year. You can only revoke it in Xentral, not via your AI application. Therefore, only share permalinks with people and tools you want to entrust with this data. You can find out how permalinks work in Xentral in the article Reports (new) - Data-driven reporting.
Tip
Have your reports displayed in the preview first and only run them once the result fits. If you have full access to reporting, you can find out your remaining credits with a question such as "How many reporting credits do I have left?".
When your AI application wants to use a tool, it usually displays the tool’s name. The following table lists what each tool is for and whether it uses credits.
Tool |
Purpose |
|---|---|
|
|
Finding saved reports by topic. Free of charge. |
|
|
Displaying the content and structure of a report. Free of charge. |
|
|
Listing collections. Free of charge. |
|
|
Listing existing schedules for exports. Free of charge. |
|
|
Displaying the remaining credits. Free of charge. |
|
|
Displaying what your plan includes in reporting. Free of charge. |
|
|
Displaying a saved report in the preview. Free of charge. |
|
|
Running a saved report. Uses credits. |
|
|
Exporting a report as an Excel or CSV file. Uses credits. |
|
|
Checking the status of an export. Free of charge. |
|
|
Displaying your own query in the preview. Free of charge. |
|
|
Running your own query. Uses credits. |
|
|
Saving a query as a new report. Free of charge. |
|
|
Creating a new collection. Free of charge. |
|
|
Changing a saved report. Free of charge. |
|
|
Setting up a recurring export by email. Free of charge. Each day the export runs uses one credit per report. |
|
|
Changing a recurring export. Free of charge. |
|
|
Creating a download link (permalink) for a report. Free of charge. The first run via the link each day uses one credit. |