For discrete and process manufacturers on Infor LN or Infor M3

Infor LN integration for AI agents: which interface reads, which one writes

Infor LN integration for an AI agent means choosing, per operation, between ION BODs and LN OData REST APIs, and for Infor M3 between MI programs and the Data Lake, all behind the ION API Gateway. Infor's own MCP servers, in limited availability since April 2026, do not name LN or M3. Pexon builds the server against your tenant.

Three ways into Infor

An agent needs all three, for different operations

LN · OData REST

The read path for LN

Infor LN publishes OData REST APIs, browsable in the Infor OS Portal under API Gateway. Good for status questions across production orders, purchase orders and warehousing, and the natural backend for most read tools.

LN and M3 · ION BODs

Where most writes belong

Business Object Documents in OAGIS XML, with Sync and Process verbs, are how Infor systems exchange changes. Write paths in LN often need a BOD rather than a simple REST call, so a tool that proposes a change publishes one, and Infor's own logic applies it.

M3 · MI programs and Data Lake

Transactions and analytics, kept apart

M3 exposes thousands of MI transactions such as CRS610MI/GetBasicData through the ION API Gateway, with terse field names like CUNO and ITNO. The Data Lake answers analytical questions in near real time, which makes it a read source, never a write path.

What Infor ships today, checked 19 September 2026

Infor announced limited availability of an enhanced Agentic Orchestrator on 22 April 2026, with MCP servers that, in Infor's words, standardise how AI models access data and take action across Infor applications and work alongside non-Infor applications. The announcement does not name LN or M3. Until Infor documents which products and versions those servers cover, an LN or M3 customer cannot plan an agent around them.

The rest is community and commercial work. For M3 there is an open-source server on the ION API, a lighter variant over Data Fabric SQL with 22 tools and one star on GitHub, and a commercial M3 server from Fortude, published in August 2025, that runs as an Azure Container App. For LN we found no specific server at all. Nothing we found handles LN's multi-company authorisation.

That is not a criticism of Infor, which is moving. It is why the useful thing to build now is a server your team owns, on the documented interfaces above, that can later sit next to Infor's own.

Why Infor is hard for agents

Six questions to answer before an agent touches LN or M3

Infor agent readiness

Honest default answers, before a blueprint

  1. For which company and site does the agent act, and who decides?Usually undefined. LN authorises per company and site, and the ION API Gateway adds its own permissions on top.
  2. Which operations must write, and does each have a BOD?Not mapped. Teams know their integrations, rarely which BOD variant an agent's write would need.
  3. Which customisations change the tables a tool would read?Known to one or two people, written down nowhere an agent could use.
  4. What throttling policy applies to a new client on the API Gateway?The default. Quota, rate smoothing and spike arrest are configurable per endpoint and rarely set for agents.
  5. Is ION API use by a non-Infor agent covered by your subscription?Unknown. Neither we nor your team can answer it; Infor can.
  6. Does anyone translate CUNO, ITNO and their siblings into words?No. Without that mapping the agent guesses from field names.

Four or more unanswered is the normal starting point. The blueprint's first week exists to turn these six into written answers.

What we build

A first tool list across LN and M3

Named operationSystemInterfaceMode
explain_order_shortageLNOData REST across production order, purchase order and warehouse dataRead: which parts are missing for an order, and why
trace_nonconformanceLNOData REST, linking an NCR to lots, serials and BOMRead: traceability for aerospace and defence quality questions
propose_supplier_date_changeLNProcess BOD for the purchase order linePropose: a person confirms before Infor applies it
get_customer_order_statusM3MI transactions for customer orders via ION API GatewayRead: order and delivery status at the order desk
query_margin_by_itemM3Data Lake SQLRead: analytical, near real time, never used to write

Every tool runs under the asking user's Infor identity where the tenant allows it, or under a narrowly scoped ION client otherwise, and each call is logged with the person, the operation and Infor's response. The final list is agreed with your Infor owners in the blueprint.

When to wait for Infor instead

If Infor confirms that the Agentic Orchestrator's MCP servers cover your LN or M3 version, and every question your agent must answer lives inside Infor, use Infor's servers and restrict them. Building a second server beside a vendor-maintained one is a dependency you would carry for nothing. We are the better choice when the agent must combine Infor with a PLM, an MES or a second ERP, or when your timeline cannot wait for a product that has no date.

Start with the Readiness Blueprint.

Two weeks, €4,900 fixed price. We answer the six questions above for your tenant, choose the interface for each operation, test whether the user's identity reaches LN's company and site authorisation, and put the subscription question to Infor. All prices are net and exclude VAT.

What we need from you: an ION administrator, a test or development tenant we may call, and the owner of the process the agent should support first. After it, a use-case pilot builds the server into your environment, and your team or ours runs it. Infor LN on the AWS European Sovereign Cloud, where Infor is a launch partner, changes where the server runs, not how it is built.

Related

Questions Infor teams ask about agents

Doesn't Infor's Agentic Orchestrator already include MCP?

It includes MCP servers, but not ones you can plan around for LN or M3 yet. Infor announced limited availability on 22 April 2026 and says its servers work across Infor applications; the announcement names neither LN nor M3. If your Infor account team confirms coverage for your product and version, and your agent only needs Infor data, use it.

Can one MCP server cover Infor LN and Infor M3?

Yes, if both run in the same Infor OS tenant. Both are reached through the ION API Gateway with OAuth 2.0 credentials, so the identity and audit layer is shared. The tools differ underneath: LN operations call OData REST or publish BODs, M3 operations call MI transactions or query the Data Lake. The agent sees one named tool list either way.

Our LN is heavily customised. Is that a blocker?

No, but it moves the work. LN's Baan heritage means the data model differs from customer to customer, so the tool list is written against your tables, sessions and BOD variants, not against a generic schema. The first week of the blueprint is mostly this mapping, and a tool that depends on a customisation names it in its description.

What does an Infor LN integration for an agent cost?

The two-week Readiness Blueprint is fixed at €4,900: interface choice per operation, the permission model across companies and sites, and a costed build plan. A first production server is usually a use-case pilot from €15,000. Whether ION API usage by a non-Infor agent affects your Infor subscription is a question for Infor, and we put it to them in the blueprint. Prices are net of VAT.

What will you not promise on Infor?

A write path that bypasses Infor's logic, and a date for anything Infor ships. Writes go through BODs or documented APIs, as proposals a person confirms. We have not built an Infor LN or M3 server for a named customer, and the community servers we checked are hobby-grade or M3-only, so we do not start from any of them.

Next step

Not a sales call. An architecture call.

Thirty minutes with the architect who would run the engagement. It helps if your ION administrator can join.