MCP, or Model Context Protocol, is a protocol for connecting models and agents to external tools, data sources, and services in a structured way.
The point is to avoid every tool integration becoming a custom one-off adapter.
An MCP server can expose capabilities such as:
read files
query databases
call APIs
search documents
inspect application state
execute domain tools
The model or agent does not directly know every API detail. It receives a structured interface describing available tools and how to call them.
Agents become useful when they can act on real systems. MCP-style integration makes tool access more standard, inspectable, and reusable.
The risk is security and permissions. Tool access must be scoped carefully. A model with broad uncontrolled tool access is not an agent; it is an incident waiting to happen.
Related: agents, context engineering, deployment.