The Model Context Protocol, explained for people who do not write code

What happened

The Model Context Protocol (MCP) is an open specification for how an AI application connects to external tools and data sources. Instead of every workflow tool writing its own connector for every service, a service can publish one MCP "server" and any MCP-aware client can use it: the same connection works in a desktop assistant, a workflow tool, and a custom Agent.

The specification covers how a client discovers what a server offers (tools, resources, prompts), how it calls them, and, in recent revisions, how authorization works for remote servers.

Why it matters

For a non-developer, two things:

What it does not change: the Agent still only does what the tools allow, and the tools still need a careful scope. MCP makes the plumbing uniform; it does not make it safe by itself.

What you can try

Nothing yet, unless your workflow tool already lists MCP servers as a connection type. If it does, treat an MCP connection exactly like any other: read the scopes, bookmark the revoke page, start read-only.

On our roadmap

This site plans to expose its library to external Agents over MCP in a later phase, so that an Agent you run elsewhere can ask our Ask AI the same questions a member can here, within the same permissions. We will write up the authorization model when it ships. We have not tested a production MCP deployment ourselves yet, which is why this piece is marked "not yet tested."