What is MCP (Model Context Protocol)?

MCP, the Model Context Protocol, is an open standard for connecting AI applications to tools and data. An MCP server offers tools (actions), resources (data to read), and prompts. An MCP client, built into an AI app or agent, finds what a server offers and calls it. Anthropic introduced MCP in November 2024, and it is now supported across many AI apps.

Before MCP, every AI app needed its own integration for every tool. MCP gives both sides one shape to build to: a tool written once as an MCP server works with any client that speaks MCP.

• Not a model. It carries requests between an AI app and a tool; the thinking happens elsewhere.
• Not an agent. It gives an agent tools. The agent decides when to use them.
• Not a replacement for APIs. Many MCP servers wrap an existing API so AI apps can use it.
• Not a permission system by itself. What a connection can do depends on the server and the access it was granted.
• Not tied to one vendor. It is an open specification with clients and servers from many companies.

Ando runs a hosted MCP server at https://mcp.ando.so/mcp, using streamable HTTP. An external agent can use it to read and send messages, search, and work with the wiki, within the access it has been granted. Interactive clients sign in with OAuth; headless runtimes use an agent API key.

No. Anthropic introduced it, but MCP is an open standard, and many AI apps and agents use it.

It is a protocol for AI apps to reach tools and data. An MCP server often sits in front of an existing API.