Glossary
Model Context Protocol (MCP)
An open protocol, introduced by Anthropic in November 2024, that standardises how AI applications connect to external tools and data sources.
Checks run on our GPUs in Switzerland. Request payloads are not stored.Security
What Model Context Protocol (MCP) means
The Model Context Protocol (MCP) is an open protocol introduced by Anthropic in November 2024. It defines how an AI application (the host, through an MCP client) talks to MCP servers, which expose capabilities in three main forms: tools the model can call, resources it can read, and prompts it can use.
Because one MCP client can connect to many servers, MCP makes it easy to give an agent a large number of tools from different vendors. That convenience also widens what an agent can do, and what can go wrong.
The MCP specification’s section on tools says that, for trust and safety, there SHOULD always be a human in the loop with the ability to deny tool invocations. In practice, asking a human about every call is not workable at scale (see approval fatigue). A pre-execution guard can triage calls so that humans see only the ones that need them. MCP Guard is itself available as an MCP server with guard_check, guard_batch and usage tools.
Related terms
How Model Context Protocol (MCP) connects to the rest of the vocabulary.