Definition
MCP (Model Context Protocol)
MCP (Model Context Protocol) is an open standard published by Anthropic in late 2024 that lets AI assistants connect to external data sources, tools and services in a uniform way — analogous to the Language Server Protocol (LSP) but for LLM context.
A typical MCP setup has a host (Claude Desktop, an IDE, a custom agent), a client per MCP server, and an MCP server that exposes resources (files), tools (functions) and prompts to the model. The security surface MCP creates is large: every server an agent connects to extends the trust boundary, and prompt injection through any of them affects the whole agent.
Why it matters
- ✓MCP servers handle real data — files, secrets, customer records — at a layer most security tools do not see.
- ✓A compromised or malicious MCP server has the same blast radius as a malicious browser extension.
- ✓Most DLP products do not yet cover MCP-bound prompts.
Common questions
Does ZeusLock cover MCP?
Yes. ZeusLock ships an MCP guard that sits between the agent and any connected MCP server, inspecting both the prompts sent and the resources returned. It is one of the few DLPs that covers MCP today.