What is an MCP Server?

What it is

An MCP server is a Model Context Protocol server that a product exposes so external AI agents can connect to its tools, data, and actions. Where a client reaches out, a server accepts connections: it publishes a set of capabilities that any MCP-compatible agent can discover and use. For an AI agent product, exposing a server means other agents can plug into it and use it as a building block.

How it works

The product runs a server that speaks the Model Context Protocol over a supported transport such as stdio or streamable HTTP. When a client connects, the server advertises its tools, resources, and prompts. The agent can then call those tools, read those resources, and receive structured results, with authentication and permissions controlled by the server. Because the interface is standardized, any MCP-compatible agent can use it without custom integration work. See our related definitions of MCP and an MCP client for the wider context.

Key capabilities

• ✓Publishes tools that external agents can call directly
• ✓Exposes resources such as documents or records for agents to read
• ✓Advertises prompts and structured actions through a standard interface
• ✓Controls access with authentication and permission scopes
• ✓Works with any MCP-compatible client without custom integration

Common use cases

• →Letting a customer's own AI assistant act inside your product
• →Making a data platform queryable by external agents
• →Exposing a CRM or ticketing system as agent-callable tools
• →Serving as a building block in a multi-agent workflow

How to evaluate one

• ?Does the vendor publish an official MCP server rather than a community build?
• ?Which tools and resources does the server actually expose?
• ?Which transports does it support, such as stdio or streamable HTTP?
• ?How are authentication and permission scopes handled?
• ?Is the server listed in the official MCP registry?

Frequently asked questions

Related definitions