What is the Model Context Protocol?
Definition
The Model Context Protocol (MCP) is an open standard for connecting AI assistants such as Claude, ChatGPT, or Gemini to outside tools and data. A product exposes an MCP server; the assistant discovers the server's tools, asks the user to sign in when needed, and calls the tools on the user's behalf during a conversation.
Servers, tools, and clients
An MCP server publishes a list of tools, each with a name, a description, and a schema for its arguments. The assistant (the MCP client) reads that list and decides when to call a tool. Tools carry annotations that say whether they only read data or also change it, so the client can ask before doing something destructive.
Signing in
Remote MCP servers usually protect their tools with OAuth. The first time you use one, the assistant sends you to the product's own sign-in page, you approve the connection, and from then on the assistant holds a token it can refresh. You can revoke that token from the product's account settings.
Where MCP is used
Claude, ChatGPT, Gemini, and Grok all accept MCP servers, as do coding agents such as Claude Code and Gemini CLI. The same server can serve all of them, which is why products increasingly publish one MCP endpoint instead of a separate integration per assistant.
How Emboss uses MCP
Emboss publishes an MCP server whose tools detect fields in a PDF, fill forms from values or documents, and batch-fill from a spreadsheet. Connect it once and any MCP-capable assistant can fill PDF forms for you.