← All courses
Course · 12 parts

Terminal to Agents

Start with a black screen you have never typed into, and end up running AI agents that reach your real tools. Twelve parts, no experience assumed, in order.

The parts are in order — start with the first one.

Part 17 Sept 2026·4 min readTerminal to AgentsCommand Line

Why Anyone Would Type Instead of Click

Part 1 of 12. You already know how to tell a computer what to do — you point and click. This is the other way, and the one reason it is worth learning. No terminal needed to read this.

Read article →
Part 28 Sept 2026·5 min readTerminal to AgentsCommand Line

Which Terminal to Open — and Why Windows Needs Care

Part 2 of 12. Pick one shell and stay in it for the whole course. On Windows the default one has commands with the right names and the wrong behaviour, which is where most beginners quietly give up.

Read article →
Part 39 Sept 2026·5 min readTerminal to AgentsCommand Line

Moving Around: pwd, ls, cd — and Four Keys So You Barely Type

Part 3 of 12. A terminal is always standing inside one folder. Once that idea lands, three commands are enough to go anywhere on your machine.

Read article →
Part 410 Sept 2026·5 min readTerminal to AgentsCommand Line

Files Without a Mouse — Make, Write, Move, and Delete

Part 4 of 12. Four commands cover almost everything you do with files. One of them deletes permanently, with no recycle bin and no confirmation, so we start there.

Read article →
Part 511 Sept 2026·5 min readTerminal to AgentsCommand Line

Make Commands Explain Themselves

Part 5 of 12. The most useful thing in this course is not a command. It is the habit of asking a tool to describe itself — after which you can read commands instead of memorising them.

Read article →
Part 614 Sept 2026·6 min readTerminal to AgentsCommand Line

Joining Tools: The Pipe, grep, and Saving the Answer

Part 6 of 12. One character connects two commands, and that is the idea the whole command line is built on. It is also why no single tool has to do everything.

Read article →
Part 715 Sept 2026·6 min readTerminal to AgentsCommand Line

When It Goes Wrong — Reading Errors Instead of Fearing Them

Part 7 of 12. People do not quit the terminal at hard topics. They quit at the first error message, which is a shame, because an error is usually the most helpful thing it prints.

Read article →
Part 816 Sept 2026·7 min readTerminal to AgentsCommand Line

Make the Work Yours — Install, Script, and Keep a History

Part 8 of 12. If you have typed the same five commands three days running, you already wrote a script. You just did not save it. Today you save it — and give the whole folder an undo.

Read article →
Part 917 Sept 2026·6 min readTerminal to AgentsConcepts

Where the Command Line Ends

Part 9 of 12. The terminal does exactly what you say, which is its greatest strength and its ceiling. Here is the honest limit of the first eight parts — and two words you need before the next three.

Read article →
Part 1018 Sept 2026·7 min readTerminal to AgentsAI Agents

An Agent in Your Terminal — and How Much to Let It Touch

Part 10 of 12. Same window, same commands. The difference is what you hand it: a task instead of a command. And it has exactly your powers, which is the part worth thinking about.

Read article →
Part 1121 Sept 2026·7 min readTerminal to AgentsMCP

What MCP Actually Is — a Shared Plug Shape

Part 11 of 12. Not a product, not a library, not a model. An agreement about shapes — and the reason it exists is a multiplication problem that used to stop integrations from ever being built.

Read article →
Part 1222 Sept 2026·8 min readTerminal to AgentsMCP

MCP or CLI? And What a Connector Actually Saves You

Part 12 of 12, the finale. The most over-built thing in AI right now is an MCP server for something curl already does. Here is how to tell the two cases apart — and the one risk worth taking seriously.

Read article →
Terminal to Agents · Ali Huzaifa