MCP server
An MCP server is the part of an app that lets AI assistants use it. It speaks the Model Context Protocol, an open standard published in 2024, and offers the assistant a list of tools, such as “list envelopes” or “start a draft”, that it can call on your behalf after you sign in.
Updated November 9, 2026 · 1 min read
On this page
How it works
The assistant is the client. You give it the server's web address, sign in to the app, and choose what the assistant may do. From then on, when your request needs the app, the assistant picks a tool, sends it the details, and reads the answer back to you. The app checks your permissions on every call, so the assistant can never do more than your account could.
An example
A roofer asks Claude which contracts are still unsigned. Claude calls the e-signature app's MCP server, gets the list, and answers in plain words, without the roofer opening the app.
In SignMaster
SignMaster runs an MCP server that Claude, ChatGPT, Gemini and Copilot connect to. Its tools look things up, start drafts and save templates; tools that would email someone only ask, and wait for your approval in SignMaster.
Questions? Answers.
Is MCP the same as an API?
It's built on top of one. An API is written for programmers; an MCP server describes its tools so an AI assistant can work out when and how to use them.
Sources
- Model Context Protocol: introduction · modelcontextprotocol.io