MCP (model context protocol)

The Model Context Protocol (MCP) is an open standard developed by Anthropic that defines a structured interface for AI models to connect with external tools, data sources, and services. MCP standardizes how a model requests access to resources (such as files, APIs, databases, or code execution environments) and how those resources return results — creating a consistent, composable layer between models and the external world.

In enterprise deployments, MCP simplifies the architecture of AI agent systems by replacing ad-hoc, custom integrations with a common protocol. Rather than building bespoke connectors for every model-tool pair, organizations can implement MCP-compatible servers for their internal systems (ERPs, OMS platforms, CRMs) and expose them to any MCP-compatible model. This reduces integration overhead, improves security through standardized permission scoping, and accelerates the deployment of AI agents that need to act on real business data and systems.