What is Model Context Protocol (MCP)?

The Model Context Protocol, or MCP, is an open standard that defines how AI applications talk to the tools and data sources they need. Before MCP, every integration between a model and an external system was bespoke, which meant work scaled with the product of clients times servers. MCP collapses that to a sum: a client implements the protocol once and can then use any MCP server, and a server implements it once and works in any MCP client. The protocol specifies how capabilities are advertised, how tools are called, how resources are read, and how transports like stdio and Streamable HTTP carry messages. Each version of the spec is dated, and servers and clients negotiate the version they share. The result is an ecosystem where documentation servers, databases, browsers, and memory layers all plug into agents through one consistent contract.