Model Context Protocol (MCP) is a shared way for AI assistants to use outside tools and information. For example, a company could let an assistant look up a report in an approved system.
The company decides what information and actions to make available. The connected system still needs to check who is asking and require approval for sensitive actions.
MCP usually connects an assistant to a separate service. WebMCP is a related draft idea for making actions on a website available to an assistant using that page.