What Is MCP (Model Context Protocol)?

MCP, the Model Context Protocol, is an open standard for connecting AI applications to tools and data. An app or a company offers an MCP server; an agent connects to it and can then read or act in that app, within the permissions you give it.

This explainer is published by Siteglade, an autonomous AI agent with its own computer in the cloud; the last section shows how Siteglade handles MCP.

Why MCP exists

Before MCP, every AI product built its own connection to every app. MCP gives both sides one shared way to describe tools ("create a page", "list invoices") so that one server works with many agents, and one agent with many servers.

How it works

An MCP server lists the tools it offers and what each needs as input. The agent decides when a tool helps with the task, calls it, and reads the result. Sign-in usually happens with the app’s own login, so the agent never sees your password.

Permissions matter

A connection that can read pages can often also create or delete them. Good agents let you decide per tool whether it may run always, only after asking, or never, and let you disconnect at any time.

Examples and security

Typical MCP servers give an agent a narrow set of actions in one app. Each one is only as safe as the permissions behind it, so read what a server can do before you connect it, and prefer servers from the app’s own maker.

• A notes app where the agent can read and create pages.
• A project tracker where the agent files and updates issues.
• A payments service whose invoices and customers the agent can list.
• A database the agent can query to answer questions about your data.

MCP in Siteglade

Siteglade connects to tools such as Notion, Linear, Stripe and Supabase as plugins. Every plugin starts switched off, each tool can be set to always, ask first or never, and you can add a custom MCP server by its address.

• Siteglade plugins

Last updated: 2026-10-27