What Is the Model Context Protocol

The Model Context Protocol (MCP) is an open standard developed by Anthropic that defines how AI agents connect to external tools, data sources, and services. It provides a structured interface for tool discovery, invocation, and result handling.

Before MCP, every agent framework implemented its own tool integration layer. LangChain had its tool abstraction. OpenAI had function calling. CrewAI had its own approach. This fragmentation meant that tool providers had to build separate integrations for each framework, and security teams had no consistent way to audit tool access.

MCP solves this by defining a standard protocol with three core concepts:

Tools are executable functions that agents can call. Each tool has a name, a description, and an input schema defined in JSON Schema format. When an agent wants to perform an action, it constructs a tool call that conforms to the schema.

Resources are data sources that agents can read. Unlike tools, resources are read-only. They provide context that helps the agent make better decisions without modifying external state.

Prompts are reusable templates that MCP servers can expose. They provide structured ways to interact with specific capabilities.

An MCP server exposes tools, resources, and prompts over a transport layer (typically stdio or HTTP with server-sent events). An MCP client connects to one or more servers and makes their capabilities available to the agent.

The security challenge with MCP is that it creates a standardized attack surface. If an agent can connect to any MCP server, it can access any tool that server exposes. This is where MCP gateways become critical. A gateway sits between the agent and MCP servers, enforcing policies about which tools can be called, with what parameters, and under what conditions.

Authensor provides an MCP gateway that applies policy evaluation, content scanning, and audit logging to every tool invocation that passes through it.