Glossary · Model Context Protocol
What is MCP (Model Context Protocol)?
MCP, the Model Context Protocol, is an open standard for connecting AI applications to outside tools and data. An AI app (the host) runs MCP clients that talk to MCP servers using JSON-RPC 2.0. Servers offer tools to call, resources to read and prompts to reuse, so one integration works across many assistants.
Also called: Model Context Protocol, MCP
MCP (Model Context Protocol), explained
Before MCP, every AI assistant needed its own plugin format. A company that wanted its data in three assistants built three integrations, each with its own authentication and its own idea of what a function call looks like. MCP replaces that with one protocol: build a server once and any compliant client can use it.
Anthropic introduced MCP as an open standard in November 2024. The specification is versioned by date; the current revision at the time of writing is 2025-11-25. It borrows its shape from the Language Server Protocol, which did the same job for code editors and programming languages.
There are three roles. The host is the AI application a person uses, such as Claude Desktop, ChatGPT, Cursor or a custom agent. Inside the host, an MCP client holds one connection to one MCP server. The server is the program that exposes capabilities. Messages are JSON-RPC 2.0 requests, responses and notifications, and the two sides negotiate which features each supports when they connect.
A server can offer three kinds of thing. Tools are functions the model may call, like create_issue. Resources are read-only context identified by a URI, like a file or a record. Prompts are reusable message templates a person picks, often shown as slash commands. Clients can in turn offer servers sampling (asking the host's model to generate text), roots (filesystem boundaries) and elicitation (asking the user a question).
MCP defines two standard transports. stdio runs the server as a local subprocess and passes messages over standard input and output. Streamable HTTP puts the server behind one HTTPS endpoint, which is what makes a remote MCP server possible. For HTTP servers, the specification's authorization section builds on OAuth 2.1.
For example, a founder connects Claude to their project tracker's MCP server. They ask, "What is overdue this week?" Claude lists the server's tools, calls the one that searches work with a due-date filter, reads the JSON result and answers in a sentence, without the founder opening the tracker.
How it works in Atrix
Atrix is an MCP server at https://atrix.dev/api/mcp, over Streamable HTTP. Claude, ChatGPT, Codex, Cursor or your own agent can read and change the workspace with the same scopes, rate limits and audit trail as the REST API, and the server offers tools, resources and prompts.
Read next
Related terms
- MCP serverAn MCP server is a program that exposes tools, resources and prompts to AI applications over the Model Context Protocol.
- MCP clientAn MCP client is the component inside an AI application that holds a connection to one MCP server.
- MCP toolAn MCP tool is a function an MCP server exposes for an AI model to call.
- Remote MCP serverA remote MCP server is an MCP server reached over the internet through one HTTPS endpoint, using the Streamable HTTP transport, instead of running on your own machine.