---
title: "Model Context Protocol (MCP)"
description: "The Model Context Protocol (MCP) is an open protocol for connecting AI applications to external tools, data sources, and reusable prompts. It was introduced by Anthropic and provides a common interface between an MCP host, such as an AI agent or development environment, and one or more MCP servers."
canonical_url: "https://promptmetheus.com/resources/llm-knowledge-base/model-context-protocol-mcp"
---

> Index of every page on this site: [llms.txt](https://promptmetheus.com/llms.txt)

The Model Context Protocol (MCP) is an open protocol for connecting AI applications to external tools, data sources, and reusable prompts. It was introduced by [Anthropic](https://www.anthropic.com/) and provides a common interface between an MCP host, such as an [AI agent](https://promptmetheus.com/resources/llm-knowledge-base/ai-agent) or development environment, and one or more MCP servers.

An MCP server can expose three primary capability types:

- **Tools:** executable operations that can be selected through [tool calling](https://promptmetheus.com/resources/llm-knowledge-base/tool-calling-function-calling).
- **Resources:** application-controlled data that can be read and included in model [context](https://promptmetheus.com/resources/llm-knowledge-base/context).
- **Prompts:** reusable templates or workflows offered by the server.

MCP uses a client-server architecture. The host manages the user experience and model interaction, while an MCP client maintains a connection to a specific server. Messages use JSON-RPC, commonly over local standard input/output or Streamable HTTP.

The 2026-07-28 revision introduced an [MCP stateless protocol](https://promptmetheus.com/resources/llm-knowledge-base/mcp-stateless-protocol) for request handling without sticky protocol sessions. It also formalized [MCP extensions](https://promptmetheus.com/resources/llm-knowledge-base/mcp-extension), including the [MCP Task](https://promptmetheus.com/resources/llm-knowledge-base/mcp-task) for long-running tool work. A [Multi Round-Trip Request](https://promptmetheus.com/resources/llm-knowledge-base/multi-round-trip-request) handles input needed during a call, while [Enterprise-Managed Authorization](https://promptmetheus.com/resources/llm-knowledge-base/enterprise-managed-authorization) supports centrally governed access.

The protocol separates capability discovery from execution. A client can inspect the tools and resources offered by a server before deciding what to make available to the model. This reduces custom integration code and allows the same server to work with multiple compatible hosts.

MCP does not make a server trustworthy. Hosts still need authentication, authorization, user consent, input validation, network restrictions, and controls around consequential actions. The [MCP authorization extension](https://promptmetheus.com/resources/llm-knowledge-base/mcp-authorization-extension) and [MCP elicitation](https://promptmetheus.com/resources/llm-knowledge-base/mcp-elicitation) address parts of that control surface, but hosts still need application-level policy. Tool descriptions and returned content can influence model behavior, so MCP integrations must also account for indirect [prompt injection attacks](https://promptmetheus.com/resources/llm-knowledge-base/prompt-injection-attack).

MCP is complementary to the [Agent2Agent Protocol (A2A)](https://promptmetheus.com/resources/llm-knowledge-base/agent2agent-protocol-a2a) and [AG-UI](https://promptmetheus.com/resources/llm-knowledge-base/agent-user-interaction-protocol-ag-ui). MCP connects an AI application to tools, context, and [MCP apps](https://promptmetheus.com/resources/llm-knowledge-base/mcp-app); A2A supports communication and task delegation between independent agents; AG-UI standardizes event streams between agents and user interfaces. Rich application surfaces may also use [Generative UI](https://promptmetheus.com/resources/llm-knowledge-base/generative-ui) or [A2UI](https://promptmetheus.com/resources/llm-knowledge-base/a2ui) when model-driven interaction needs a structured UI contract.

Discovery can also happen outside the protocol connection itself. For example, an [agents.txt](https://promptmetheus.com/resources/llm-knowledge-base/agents-txt) file can advertise agent-facing endpoints, while the MCP connection still governs concrete tool and resource execution.

The current specification and SDKs are available on the [Model Context Protocol website](https://modelcontextprotocol.io/docs/2026-07-28/getting-started/intro).

## Sitemap

See the full [sitemap](https://promptmetheus.com/sitemap.md) for all pages.