Skip to content
The State of Exposed AgentsRead the report
Guard0

What is the Model Context Protocol (MCP)?

The Model Context Protocol (MCP) is an open protocol that standardizes how AI applications connect to external tools and data. An MCP server exposes tools, resources and prompts. An MCP client inside an assistant or agent discovers them and calls them on behalf of the model, over standard input and output or over HTTP.

Also called MCP, MCP server. Last reviewed .

Why it matters

MCP made it easy to give an agent a new capability: add a server, and the model can call its tools. The specification is explicit that tools are controlled by the model, and says that "there SHOULD always be a human in the loop with the ability to deny tool invocations."

Every MCP server is also a new integration to account for. It runs code, often holds a credential, and describes its own tools to the model in text the model will read. A team that does not know which servers its agents connect to does not know what those agents can do.

MCP security
MCP security is the practice of controlling the risks MCP servers and clients add.
Tool poisoning
Tool poisoning abuses the tool descriptions an MCP server sends to the model.
Confused deputy
The MCP specification names the confused deputy problem as a risk for MCP proxy servers.
Agentic AI
Agentic AI is the kind of software that uses MCP to reach tools.

How Guard0 handles it

Guard0 lists every MCP server your agents connect to, local and remote, with the client that uses it, the tools it exposes and the credential it holds.

MCP security with Guard0

Sources

  1. Model Context Protocol specification. Model Context Protocol.
  2. MCP specification: Tools. Model Context Protocol.
  3. MCP specification: Security Best Practices. Model Context Protocol.

Find the agents in your own company.

Start free: discover the AI on three people's machines, register one AI system and connect one repository or AWS account.