What Is MCP (Model Context Protocol)?
Also known as: Model Context Protocol, MCP Server
Definition
MCP (Model Context Protocol) is an open standard that defines how AI applications connect to external tools and data sources, so any compliant client can use any compliant server without custom integration code for each pair.
MCP (Model Context Protocol) Explained
MCP addresses a combinatorial problem. Without a standard, connecting five AI applications to eight enterprise systems means writing and maintaining forty bespoke integrations. MCP defines a client-server contract instead: each system is exposed once by an MCP server, and any MCP-capable client can consume it. The integration count collapses to eight servers plus five clients, and each side evolves independently.
The protocol defines a small number of primitives. Tools are callable functions with described parameters that the model can invoke to take an action or run a query. Resources are readable data the client can pull into context, such as a file or a record. Prompts are reusable templates a server can offer. Servers advertise their capabilities on connection, so a client discovers what is available at runtime rather than having it hardcoded, which is what makes the ecosystem composable.
For ERP work the natural pattern is one MCP server per system boundary. A SyteLine server exposing bounded operations for orders, inventory, and job status. An Infor LN server for its own domain. A document server over your controlled procedure library. Each server owns its own authentication, translates between the protocol and the system's native API or database, and enforces which operations are exposed at all, which is a meaningful containment boundary.
Security deserves attention proportional to what the servers expose. An MCP server is a privileged bridge into a system of record, so it should authenticate the calling context, propagate end user identity to the underlying system rather than acting through a shared account, expose the narrowest useful set of operations, and log every invocation. In on-prem and air-gapped deployments servers typically run inside the same enclave as the model, communicating over local transport with no external dependency.
Why It Matters
- Replaces N times M bespoke integrations with one server per system, collapsing the long-term maintenance burden of AI connectivity.
- Lets you swap the model or the AI client without rewriting integrations, protecting the investment as the model landscape shifts.
- Server boundaries define exactly which ERP operations AI can reach, making the exposed surface reviewable and explicit.
- Runs entirely inside an enclave over local transport, so it fits on-prem and air-gapped architectures without external dependencies.
In Practice
A common design mistake is building one MCP server that exposes a generic run-SQL tool against the ERP database. It is fast to build and impossible to secure, because the exposed surface is the entire schema. Expose named business operations with validated parameters instead, so the reviewable surface is a short list rather than the whole database.
Frequently Asked Questions
How is MCP different from a normal REST API?
A REST API is designed for programmers who read documentation and write client code. MCP is designed for AI clients that discover capabilities at runtime, so servers advertise their tools, parameters, and descriptions in a machine-readable form the model can reason about. MCP servers commonly wrap existing REST APIs rather than replacing them, adding discoverability and a consistent contract.
Can MCP servers run in an air-gapped environment?
Yes. MCP servers are ordinary processes that communicate with the client over local transport such as standard input and output or a local socket. Nothing in the protocol requires internet access. In an air-gapped enclave the model, the MCP servers, and the systems they front all run inside the boundary, with dependencies staged in through your normal transfer process.
Related Terms
AI Agent
An AI agent is a system that wraps a language model with tools, memory, and a goal so it can plan a sequence of steps, call external systems, observe the results, and iterate until the objective is met or a limit is reached.
AI Guardrails
AI guardrails are the technical and policy controls placed around an AI system that constrain what it may receive as input, what it may produce as output, and what actions it is permitted to take, independent of what the model itself decides.
On-Prem AI
On-prem AI is the practice of running artificial intelligence models on hardware you own and control, inside your own data center or plant, so prompts, documents, and model weights never leave your network.
Go Deeper
Business Process Automation ROI Calculator
Estimate annual savings, payback period, and three-year ROI from automating repetitive business processes with AI and workflow automation.
Build vs Buy: AI Agents for the Enterprise
Build vs buy AI agents for the enterprise: total cost, timelines, and risk compared for manufacturers, with a decision framework CIOs can use today.
On-Prem LLM Deployment Architecture: Reference Guide
Reference architecture for on-prem LLM deployment: inference servers, GPU sizing, RAG pipelines, and security zones for regulated manufacturers.
Working with MCP (Model Context Protocol) in a live environment? Our engineers do this every day - and our AI agents automate most of it.