A8gent
GlossaryAgent concepts

What is MCP?

Also called: Model Context Protocol, MCP server

MCP, the Model Context Protocol, is an open standard for connecting AI applications to tools and data sources, so one integration can be reused by any compatible AI client instead of being rebuilt per app.

Written by A8gent
Contents
  1. At a glance
  2. What MCP is
  3. What a server exposes
  4. Limits, security and cost
  5. What you should know
  6. FAQ
  7. Related terms

At a glance

Category
Agent concept - interoperability
Full name
Model Context Protocol
Introduced
By Anthropic in November 2024, later donated to a neutral foundation
Primitives
Tools, resources and prompts
Transports
Local standard input/output, or HTTP for remote servers
Does not solve
Tool overload - every exposed tool still consumes context on every call

What MCP is

The Model Context Protocol is an open standard for connecting AI applications to external systems. Anthropic introduced it in November 2024 and later donated it to a neutral foundation; it is now implemented by a wide range of AI clients and thousands of servers. The specification describes a client-server architecture: an MCP server exposes capabilities, an MCP client inside an AI application consumes them.

The problem it solves is combinatorial. Without a standard, every AI application needs its own integration with every system - N applications times M systems. With one, a system is wrapped once as a server and any compatible client can use it. It is the same argument that produced the Language Server Protocol for code editors, and the comparison is apt: MCP is plumbing, not intelligence.

What a server exposes

The specification defines three primitives. Tools are functions the model can invoke - create a job, look up a customer - and they map onto the same tool calling mechanism the model already uses. Resources are data the application can read into context, like a file or a record. Prompts are reusable templates the server offers for common tasks.

Transport is either standard input and output for a server running locally, or HTTP for a remote one. That distinction matters commercially: a local server runs on the user's machine with the user's credentials, while a remote server is a service you operate and must authenticate, authorise and monitor like any other API.

Limits, security and cost

MCP does not make the model better at choosing tools. Every tool an MCP server exposes still lands in the context window as a definition on every call, so connecting six servers with ten tools each produces the same overload and mis-selection problems as declaring sixty functions by hand. Connect the servers a given agent actually needs.

Security deserves care. An MCP server is code you are granting access to your systems, and a third-party one is a supply-chain dependency; a server that reads untrusted content is also a prompt injection surface. Treat servers like any other dependency: prefer official ones, pin versions, scope credentials narrowly, and keep human approval in front of destructive tools. The specification includes authorization requirements for HTTP transports - read the version you are implementing, since the spec is dated and revised.

What you should know

For a business, the practical value of MCP is reuse and portability. Wrap your CRM, your job management system and your document store once, and any agent you build later - in n8n, in a desktop client, in your own code - can use the same servers. That reduces the cost of the second and third automation dramatically, and it makes switching AI clients a configuration change rather than a rewrite, which is a real answer to vendor lock-in.

If you are choosing between building direct function calling and standing up MCP servers, the deciding question is how many consumers the integration will have. One agent, one integration: skip it. Several agents or several clients: MCP pays for itself. See MCP vs function calling for the full comparison.

MCP FAQ

What is MCP in AI?

The Model Context Protocol, an open standard for connecting AI applications to external tools and data. A system is wrapped once as an MCP server, and any compatible AI client can then use it, instead of every application building its own integration with every system.

What is the difference between MCP and function calling?

Function calling is how a model requests a tool. MCP is how tools are packaged and discovered so they can be reused across applications. They are complementary: an MCP client surfaces the server's tools to the model as ordinary tool definitions, and the model calls them the usual way.

What does an MCP server actually expose?

Three things: tools, which are callable functions; resources, which are data the application can read into context; and prompts, which are reusable templates for common tasks. Most business servers use tools heavily and resources occasionally.

Is MCP secure?

The protocol defines authorization requirements for remote transports, but the risk is what you connect. A server has real access to your systems, a third-party one is a supply-chain dependency, and any server returning untrusted content is an injection surface. Scope credentials narrowly and keep approval on destructive tools.

Do I need MCP for a single automation?

Usually not. One agent with a handful of bespoke tools is simpler to build with direct function calling. MCP earns its place when several agents or several different AI clients need the same integrations, or when you want the tool layer maintained separately from the agents.

Does connecting more MCP servers make my agent better?

No, and often the opposite. Every tool from every connected server occupies context on every call and adds another option for the model to choose wrongly. Connect only the servers a given agent needs for its job, exactly as you would keep a hand-written tool list short.

Primary source

Model Context Protocol - specification

Checked August 2026. Anything priced or rate-limited moves, so verify against the source before you budget.

Want this built rather than explained?

Knowing what mcp means is one thing; having it running in your business is another. We scope one workflow at a time, quote a fixed price in writing after a free consult, build it, and hand it over with documentation you own.