Model Context Protocol (MCP) explained: An FAQ
By Steven Van ·
Vercel's FAQ explains MCP, the open spec that lets developers build one server to give AI models controlled access to their app's data and tools.
Vercel published an FAQ on the Model Context Protocol (MCP), an open standard for connecting AI models such as Claude or GPT-4 to external tools and systems: an app's API, a product database, a codebase, or a desktop environment. Vercel describes it as a spec rather than a library or SDK, similar in kind to REST or GraphQL but built for AI agents.
The point of MCP is to let a model reach beyond its training data when it hits the limits of its own knowledge: instead of guessing, it calls a defined function and gets real, current data back. Building an MCP server means developers write an integration once and reuse it across AI platforms, rather than building custom connections for each one.
According to Vercel, teams would build an MCP server to give a model controlled access to their own app or service, such as placing an order, fetching user data, or running a search, with the exact capabilities and rules defined by the developer rather than inferred from API docs. For a single app or a single AI provider, Vercel notes it may be simpler to define tools directly in the LLM call instead.