Browse docs
CLI · MCP · API
The three ways humans, software, and AI talk to software — and how the Model Context Protocol lets AI actually use them. A visual, source-backed overview.
The Big Picture: Three Ways to Talk to Software
A CLI is for humans. An API is for software. MCP is for AI. Same idea — a contract for asking a system to do something — aimed at three different callers.
CLI: Commands, Pipes, and the Unix Idea
The oldest door, and still the fastest. A command-line interface turns plain typed instructions into work — and a single idea from 1970s Bell Labs still shapes how we automate today.
API: How Software Talks to Software
An API is a promise: send a request shaped like this, get a result shaped like that. It is the contract that lets two systems cooperate without knowing each other's insides.
Authentication: The Lock on Every Door
Every door you have met so far assumed the system trusts whoever is knocking. In the real world it cannot. This lesson is the lock: how a system proves who you are, decides what you may do, and carries the answer on every request.
MCP: Giving AI Real Tools
If an API is how software talks to software, MCP is how AI talks to everything. This lesson builds it from the ground up — the problem it solves, how it works, the three things a server can offer, and exactly what happens when an agent uses one.
The MCP Ecosystem: Real Servers and Clients
The last lesson explained what MCP is. This one answers the question that comes right after: what is actually out there? The quiet truth is you rarely build a server — thousands already exist, and connecting one is mostly config.
Webhooks: When Software Calls You
Every door so far works one direction: you knock, the system answers. But some of the most important things in software are events you cannot predict. Webhooks flip the direction — instead of you calling the server, the server calls you.
Putting It Together: One Capability, Three Doors
The point of learning all three is seeing they are the same capability behind different doors. Build the logic once; open a CLI for people, an API for software, and an MCP server for AI.