The Model Context Protocol (MCP) is an open standard for connecting AI agents to external tools and data through a common interface, so any compatible client can use any compatible server.
Before MCP, every agent needed bespoke integrations for each tool it used. MCP replaces that with one protocol: a server describes the tools it offers and how to call them, and any MCP-aware client can discover and invoke them. It's the reason a single GitHub or Postgres server works across Claude Code, Cursor and other clients without rewriting the integration.
Practically, MCP defines how tools, their inputs and outputs, and results are exchanged between client and server, usually over stdio locally or HTTP for hosted services. That standardisation is what lets an ecosystem of interchangeable servers exist at all.
It's an open protocol adopted across the agent ecosystem; servers and clients from many vendors interoperate through it.
No. It's an open standard, any client can implement it, and many do.