---
title: "Model Context Protocol (MCP)"
description: "The Model Context Protocol (MCP) is an open-source standard for connecting AI applications to external systems such as data sources, tools and workflows."
canonical: "https://agentdrive.sh/glossary/mcp"
---
# Model Context Protocol (MCP)

> For the complete documentation index, see [llms.txt](https://agentdrive.sh/llms.txt). Every docs page is also served as markdown at its URL plus `.md`, and the full docs are in one file at [llms-full.txt](https://agentdrive.sh/llms-full.txt).

> The Model Context Protocol (MCP) is an open-source standard for connecting AI applications to external systems such as data sources, tools and workflows.

Source: https://agentdrive.sh/glossary/mcp. Part of the AgentDrive glossary: https://agentdrive.sh/glossary.

MCP uses a client-server architecture: a host application, such as Claude Code, creates one MCP client for each MCP server it connects to. Messages use JSON-RPC 2.0 over stdio for a local server process or Streamable HTTP for a remote one. Servers expose three core primitives: tools the model can call, resources that provide context, and prompts that structure interactions.

## In AgentDrive

MCP is AgentDrive's main integration surface. The `@adrive/mcp-server` package runs locally over stdio, and a hosted endpoint at https://mcp.agentdrive.sh/mcp serves remote clients such as claude.ai and ChatGPT. Any MCP client can use the workspace tools.

## Related terms

- [MCP server](https://agentdrive.sh/glossary/mcp-server/index.md): An MCP server is a program that provides context to AI applications over the Model Context Protocol, exposing tools, resources and prompts that a connected MCP client can discover and use.
- [Coding agent](https://agentdrive.sh/glossary/coding-agent/index.md): A coding agent is an AI tool that works through a software task in a code repository: it reads and edits files, runs commands such as tests, and reports back.
- [Principal (AgentDrive)](https://agentdrive.sh/glossary/principal/index.md): In AgentDrive, a principal is the identity behind an API key: a human or a team that owns workspaces, holds the plan and is billed, whichever agents act on its behalf.

## Go deeper

- [MCP tools reference](https://agentdrive.sh/docs/mcp-tools)
- [Integrations](https://agentdrive.sh/integrations)

## Official source

- [modelcontextprotocol.io: What is MCP?](https://modelcontextprotocol.io/docs/getting-started/intro)

