MCP (Model Context Protocol)
What is MCP (Model Context Protocol)?
MCP (Model Context Protocol) is an open standard defining how AI agents and assistants connect to external tools, data sources and services through one shared interface, instead of a custom connector for every integration. AdminKit's MCP gateway sits between an agent and the MCP servers it calls, applying the protocol's own security guidance to that traffic.
By AdminKit team, Netbase JSC · Updated
MCP (Model Context Protocol) is an open standard that defines how AI agents and assistants connect to external tools, data sources, and services through a common interface, instead of every integration needing its own custom, one-off connector. Anthropic introduced and open-sourced MCP on November 25, 2024. Every message between an MCP client and server follows JSON-RPC 2.0. An MCP server exposes a set of tools or resources an agent can call; an MCP client, often an AI application, discovers and invokes them following the protocol’s shared conventions for authentication, request/response shape, and capability discovery. Because MCP servers are increasingly exposed to untrusted or semi-trusted agents, the protocol’s own specification includes security considerations — like avoiding naive token passthrough — that implementers have to actively follow rather than assume are automatic. Two of its primitives carry their own consent requirements for this reason: a prompt is surfaced to the user rather than invoked by the agent directly, and a sampling request must get explicit human review and approval before the completion it returns is used. A2A (Agent2Agent Protocol) addresses a related but distinct problem — one agent delegating a task to another agent, rather than an agent calling a tool.
Browse the AdminKit glossary index for other related terms.
Frequently asked questions
What messaging format does MCP use?
Every message between an MCP client and server follows JSON-RPC 2.0, the same request/response/notification format MCP builds its own authentication and capability-discovery conventions on top of.
Why does MCP need its own security considerations?
Because MCP servers are increasingly exposed to untrusted or semi-trusted agents, so the protocol's specification includes guidance -- like avoiding naive token passthrough -- that implementers have to actively follow rather than assume is automatic.
Which MCP primitives require explicit human consent?
A prompt is surfaced to the user rather than invoked by the agent directly, and a sampling request must get explicit human review and approval before the completion it returns is used -- both carry their own consent requirements for exactly this reason.
How is MCP different from A2A (Agent2Agent Protocol)?
MCP addresses an agent calling a tool or data source; A2A addresses a related but distinct problem -- one agent delegating a task to another agent.