









































Works out-of-box with your favorite AI agents

Sage X3 is a manufacturing and distribution ERP from Sage for mid-market and larger companies, covering finance, inventory, production, purchasing, and multi-site sales. With Makini's MCP server, AI agents connect to Sage X3 data and workflows from day one, without writing any integration code.
Have any questions? We’re here to help You
Yes, Makini provides an MCP server for Sage X3: you authorize your Sage X3 login through Makini Link and paste the endpoint Makini issues into Claude, ChatGPT, Cursor or any MCP client. Sage X3 holds customers, suppliers, products, sales and purchase orders, stock and work orders, organized by company and site, and Makini's unified models include sales orders, purchase orders, stock items, production orders and sites. get_connection_details shows which objects and fields your connection exposes.
Sage X3 tracks stock per site, by lot and location, and each quantity carries an A (accepted), Q (in quality control) or R (rejected) status. An agent can use this to separate usable stock from quantities still in quality control before it confirms an order or suggests a purchase. Sub-statuses are set up in each installation, so confirm their meaning with your team first.
In Sage X3 a work order is an order to make one or more products, with materials taken from the bill of materials and operations from the routing. An agent can list firm and planned work orders for a production site, compare their material needs with stock, and flag shortages to planners. Makini's unified models include production orders, BOMs and routings, subject to what your connection exposes.
Makini's MCP tool list includes create_purchase_order and create_sales_order, but whether a write succeeds depends on what your Sage X3 connection exposes and on the connecting user's rights. In Sage X3, user authorizations decide which sites and data a user can reach, and the agent can do only what those credentials allow. Test writes against a non-production folder first.
Sage offers Sage X3 as a fully managed SaaS service and also supports private cloud and on-premises installations. For a server on your own network or in a private cloud, IT may need to allowlist Makini's fixed IP addresses in the firewall, or set up a VPN if allowlisting is not possible. The connection itself is then authorized through Makini Link with credentials for your Sage X3 instance.
Sage X3 offers SOAP web services on its classic functions, a REST Web API built on representations, and a newer GraphQL API. Which one fits depends on the module, since not every module has been moved to representations yet. Your agent does not call these APIs itself: it talks to Makini's MCP endpoint, and Makini's hosted connector handles the Sage X3 side.
Decide which folder the agent should reach, since Sage X3 web service endpoints point to a specific folder, such as a live or a test folder. Use a dedicated Sage X3 user whose authorizations cover only the companies, sites and functions the agent needs. Also check authentication: for web services, basic authentication works only on-premises, and OAuth 2 is the only method in Sage X3 Online.
Yes, one agent can report across several Sage X3 companies and sites within the user's rights. Sage X3 supports several companies and sites in one folder, with multiple currencies and legislations, and each site belongs to a company. An agent can compare open orders or stock across sites in a single question, as long as the connecting user is authorized for each site. Check the currency on each document before it totals amounts across companies.