What is MCP in manufacturing?

It is the use of the Model Context Protocol to connect AI assistants and agents to industrial systems, such as the MES, ERP, QMS or SCADA, through a standardized interface.

What is an MCP server for MES?

It is a tool that exposes the data and actions of a Manufacturing Execution System to MCP-compatible AI applications, such as Claude or ChatGPT.

Does MCP replace the MES?

No. The MES continues to manage production execution and remains the official record. MCP is simply the way AI accesses that data.

Does MCP replace MES-ERP integrations?

No. Integrations between systems still exist. MCP connects AI to the systems; it does not connect the systems to each other.

Can AI change production data through MCP?

Only if the MCP server exposes actions with that permission. In industrial and regulated environments, it is recommended to start with read-only access and require human confirmation for any action.

Is MCP compatible with GxP and 21 CFR Part 11?

The protocol itself is neither compliant nor non-compliant. Compliance depends on the implementation: access control, audit trail, data integrity and inclusion in the system validation.

Does plant data leave the company when using MCP?

It depends on the architecture. An MCP server can run on the company's own infrastructure, but the data the AI queries is sent to the AI model being used, so the choice of assistant and its data processing terms should be assessed.

Where should a factory start with MCP?

With a read-only pilot on a limited set of data, such as production indicators for one line, with well-defined users and permissions.

It is the RAILES solution, announced at the 2026 Manufacturing Tech Summit, for making RAILES MES data accessible to AI models through the Model Context Protocol.

MCP in Manufacturing | Connecting AI to MES Production Data

Summary

The Model Context Protocol (MCP) is an open source standard that lets AI assistants and agents query data and take actions in external systems through a common interface. In manufacturing, this means connecting AI to the systems where production data lives: the Manufacturing Execution System (MES), the ERP, the quality system, maintenance and sensors.

With an MCP server on top of the MES, a production manager can ask in natural language about the status of an order, the deviations on a batch or the efficiency of a line, and get an answer based on the plant's real data. In regulated industries such as pharmaceuticals, this access must follow the same data integrity, access control and traceability rules that apply to the MES itself.

RAILES, a Portuguese MES company, announced RAILES MCP at the 2026 Manufacturing Tech Summit, with the goal of making MES data accessible to AI models.

What Is MCP?

MCP is an open protocol that defines how an AI application connects to external data sources, tools and workflows. The official documentation compares it to a USB-C port for AI applications: a standardized connection that works with many different systems.

It was introduced by Anthropic in November 2024 and is now an open source project under the Linux Foundation. It is supported by assistants such as Claude and ChatGPT, and by tools such as Visual Studio Code and Cursor.

MCP is built on three parts:

Host: the AI application the user works with, such as a chat assistant.
MCP client: the component of the host that maintains the connection to a server.
MCP server: the program that exposes a system's data and actions, for example an MES.

An MCP server can offer three types of capabilities: resources (data to query), tools (actions the AI can request to run) and prompts (reusable interaction templates).

Why Does MCP Matter in Manufacturing?

The value of AI in a factory depends on access to current, contextualized production data, and that data is spread across several systems. Planning lives in the ERP, shop-floor status in the MES, quality records in the QMS, maintenance in the CMMS, and measurements in sensors and SCADA.

Connecting every AI tool to each of these systems through custom integrations creates what MESA International describes as "N×M" complexity: every new source or tool needs its own integration, which is hard to maintain and to audit. MCP solves this with a single interface: each system is exposed once through an MCP server, and any compatible assistant can use it.

For manufacturers, this brings three concrete benefits:

Natural language access: decision-makers on the plant floor can query data without relying on custom reports or requests to the IT team.
Real context for the AI: answers are based on MES data, not on the model's generic knowledge.
Centralized governance: AI access goes through a controlled point, with permissions and a log of operations.

How Does MCP Connect an AI Assistant to an MES?

The AI assistant communicates with an MCP server, and it is the server that queries the MES and returns the answer. The assistant never accesses the MES database directly: it only sees what the MCP server exposes, with the permissions that have been defined.

The MCP server can run on the plant's own infrastructure or as a remote service. For remote servers, the MCP specification recommends OAuth authentication. If the plant has other systems, such as ERP, QMS or CMMS, each one can have its own MCP server, and the assistant combines context from several servers in the same answer.

What Are the Use Cases for MCP in a Factory?

The most immediate use cases are queries: asking the assistant what is happening in production and getting an answer based on MES data. The examples below are illustrative and depend on what each MCP server exposes.

Industry	Example question to the assistant	Data involved
Pharmaceuticals and biotech	"Which batches from this week still have open deviations before release?"	Electronic batch records (eBR), deviations, review status
Automotive	"Which parts from this lot went through the machine that had a stoppage yesterday?"	Genealogy and traceability, equipment events
Packaging	"What was the efficiency of line 2 on the last shift, and where was the most time lost?"	OEE, stoppages, downtime reasons
Aerospace and defense	"Which operations on this order still have no inspection recorded?"	Work orders, routings, inspection records
Plastics	"Which process parameters were active when the reject rate went up?"	Machine parameters, quality records

Beyond queries, an MCP server can expose actions (tools), such as opening a maintenance request or preparing a draft shift report. MESA recommends a phased adoption: start with read-only access, move to agents that recommend without executing, then to controlled actions with human review, and only at the end, once validated, to closed-loop automation.

What Security Measures Does MCP Require in a Factory?

An MCP server gives AI models access to data and, in some cases, to actions on production systems, so it should follow the same security principles as any industrial integration. The official MCP best practices and MESA point to five rules:

Least privilege. Each user and each assistant only accesses what they need; additional permissions are requested when an operation requires them.
Authentication and authorization. Remote servers should use OAuth and never accept tokens issued for other services.
Human confirmation for actions. Any action with an impact on production or quality should be confirmed by a person.
Logging of operations. All queries and actions made through MCP should be logged, identifying who requested them.
Trusted server sources. Only MCP servers from known vendors should be installed, and local servers should run with minimal privileges.

In plants with industrial cybersecurity programs, such as those based on the ISA/IEC 62443 standard, the MCP server should fall under the same network segmentation and access control as the other IT and OT systems.

What Is the RAILES Approach to MCP?

For RAILES, AI in the factory is only useful when it is built on reliable, governed and contextualized production data, and the MES is the system that generates it. RAILES is a Portuguese Manufacturing Execution System company, working with the pharmaceutical, automotive, packaging, aerospace and defense, and plastics industries.

The RAILES approach to MCP rests on three components:

Component	Role
RAILES MES	Manages production execution and generates the source data, with native electronic batch records (eBR), audit trail and electronic signatures.
RAILES Data Fabric	Brings MES data and data from other plant systems, such as ERP, SCADA or CMMS, into a single source of truth.
RAILES MCP	Makes MES data accessible to AI models through the Model Context Protocol.

RAILES MCP was announced at the Manufacturing Tech Summit on 22 October 2026, in Lisbon. The logic is that of the protocol itself: production data becomes available through an open standard, to the AI assistants the company already uses, without custom integrations.

Because the underlying data comes from an MES designed for regulated environments, the AI works on data that already has an audit trail and traceability, an essential requirement in industries such as pharmaceuticals.