MCP (Model Context Protocol)
The Model Context Protocol (MCP) is an open standard, originally released by Anthropic in late 2024, that defines how AI agents connect to external tools and data sources. It uses JSON-RPC 2.0 over standardized transports so that any MCP-compatible agent can discover and invoke any MCP-compatible server without custom integration code.
Why It Matters
Before MCP, every combination of AI agent and external service required a bespoke integration. Connecting Claude to GitHub was a different engineering problem than connecting Cursor to GitHub, even though both needed the same capabilities. MCP replaced this with a universal protocol: build an MCP server once, and every compatible client can use it.
MCP servers expose three primitives: tools (functions the agent can invoke), resources (data the agent can read), and prompts (templates that structure how the agent interacts with the server). The agent's host application manages the context window and decides when to call tools based on the user's request.
By mid-2026, MCP has been adopted by Anthropic, OpenAI, Google DeepMind, and Microsoft. The public registry lists thousands of community servers. The protocol's July 2026 release candidate introduces a stateless core, making it deployable on ordinary HTTP infrastructure without persistent connections.
For security teams, MCP's standardization is a double-edged property. It makes agent-tool connectivity frictionless, which accelerates adoption. It also makes governance possible: because all agent-tool traffic flows through a known protocol, a gateway can intercept, evaluate, and log every interaction.