Request a call

Model Context Protocol (MCP)

Definition

The Model Context Protocol is an open standard for connecting AI applications to outside tools and data, so a single integration works across compatible assistants and agents.

Updated

How it works

Anthropic released MCP in November 2024, and in December 2025 donated it to the Agentic AI Foundation, a fund under the Linux Foundation. ChatGPT, Claude, Gemini, Microsoft Copilot, Cursor and Visual Studio Code support it.

The protocol has 3 roles, as the Model Context Protocol documentation sets out. The host is the AI application. It opens an MCP client connection to each MCP server. A server wraps a system, such as your CRM, a database or a file store, and exposes it as tools (actions the model can call), resources (data it can read) and prompts (reusable templates). Messages use JSON-RPC 2.0. Local servers usually talk over standard input and output; remote servers use Streamable HTTP, with OAuth recommended for sign-in.

Before MCP, each assistant needed its own connector for each system. Now a server built once can serve any compliant host.

How businesses use it

  • Giving staff assistants such as Claude or ChatGPT controlled access to internal systems, for example read-only access to the ticket queue
  • Letting a custom AI agent use the same connectors your staff tools use
  • Publishing an MCP server for your own product, so customers' assistants can work with it
  • Replacing a pile of one-off integrations with a small set of maintained servers

Many SaaS vendors now publish official MCP servers. Where they do not, we build them as part of workflow integrations.

Where it goes wrong

MCP standardizes the connection and leaves security to you. A server with broad write access lets any connected model act with those permissions, and text inside a tool result can carry prompt injection. Third-party servers from unknown publishers are code running with your credentials.

Treat every server like any other integration: least-privilege scopes, separate read and write tools, approval on anything destructive, audit logs and a named owner. Also check whether you need MCP at all. A scheduled sync in n8n that moves records between 2 systems needs no model and no protocol.

Related terms

  • Function calling

    Function calling lets a language model ask your software to run a named function, such as a CRM lookup, by returning the function name and arguments as JSON.

  • AI agent

    An AI agent is a language model that chooses and calls tools in a loop, reading each result, until it finishes a task or hands off to a person.

  • Prompt injection

    Prompt injection is an attack where text given to a language model, typed by a user or hidden in a document, overrides the instructions the system was built with.

  • n8n

    n8n is a workflow automation platform, self-hosted or cloud, that connects apps and APIs in visual workflows and supports custom code and AI agent steps.

  • AI guardrails

    AI guardrails are checks placed around a language model that block, rewrite or escalate inputs and outputs that break your rules on safety, accuracy, privacy or scope.

All 30 definitions are in the AI and search glossary.

Find out where this fits in your own systems

Bring one process or page to a 30-minute call.

Request a 30-minute scoping call

We reply within 1 working day and suggest times that suit you.