Connect Claude to the Tools You Already Use

You are still the transport. Something comes out of your calendar by hand, goes into Claude by hand, and the answer goes back out the same way, six times a day.

If you have opened the connectors menu once and closed it again because the next word on the screen was an acronym, you are not the only one.

There are exactly two things to understand here, and you only ever touch one of them. That is the whole topic. It is a great deal smaller than its vocabulary makes it look.

One. A connector is approved access to one app.

You give Claude permission to reach a specific tool you already use. Your calendar, your email, your documents, your notes. From then on it can read what is in there and use that as the input to a job you described.

Approved is the load-bearing word. Nothing connects without you saying so, one app at a time, and you can withdraw it whenever you like.

A postal carrier can reach your mailbox, read the address on an envelope and put it where it belongs. They do not own your house. They have narrow access to one spot, for one purpose, granted by you. That is the shape of a connector, and it is worth holding onto, because it is also the shape of what can go wrong.

Reading is the gentle capability, and most of the value is already sitting in it. A job that reads your calendar and writes you a summary changes nothing at all. It looks and it reports. Nothing was sent, nothing was moved, and the worst version of it is a paragraph you ignore.

Two. MCP is the socket, and you never touch it.

Sooner or later you will hit the letters MCP, and they sound like homework. They stand for Model Context Protocol, three words to note once and then forget. It is the shared standard that lets tools plug into Claude without somebody building a custom bridge for each pairing.

Think of a wall socket. Any lamp, any charger, anything with the right plug fits the same hole. Nobody rewires the wall for each new device. MCP is that socket, which is the entire reason connecting a tool is a menu item rather than a project.

You approve the app and describe what you want in plain words. The handshake underneath is Claude's problem. Knowing the term is useful for one reason only: the next time you see it in a menu or an article, you will know it means the standard that lets tools connect, and you will not close the tab.

Connect the least you can get away with.

The habit worth forming on day one is least access. If the job reads your calendar, connect the calendar and nothing else.

There are two reasons and the second is the one people miss. The obvious one is that fewer connections mean a smaller surface to keep an eye on. The quieter one is that a narrow connection produces better answers. Give Claude four sources for a job that needs one and you have handed it three extra ways to answer a question you did not ask.

You can always add another when a job actually needs it. Adding takes a minute. Working out which of your nine connected apps produced a strange answer takes an afternoon.

Free Claude toolkit. The companion pack for my Claude book is free: ready-made Project instruction templates, and more than five hundred numbered prompts you can copy and paste. No card, and you can leave the list in one click. Get the toolkit →

Start from a job, not from the menu.

The common failure here is connecting things because you can, and then waiting for a use to appear. It does not appear. You end up with six approvals and the same manual copying you had before.

Work the other way round. Name one thing you currently do by hand every week. Write down which single app holds the information for it. Connect that one, and then describe the job in the same plain brief you would write for anything else.

The first three worth building are almost always the same three. A short digest of what today holds. A queue of drafted replies for one recurring kind of message. A summary of the long thing that lands in one folder each week. Every one of them reads and prepares. Not one of them commits to anything.

What this will not do.

A connector does not give Claude judgement about your information. It gives it access. What comes back is only as good as the app is tidy, and connecting a chaotic inbox gets you a well-written summary of chaos.

It also does not put reading and acting on the same footing. Reading is safe. Anything that sends, deletes or buys sits under a completely different rule, and what not to let run unattended is where I set that one out.

And it will not keep looking like this. The menu moves, the names change, the list of available apps grows every few months. The two ideas underneath do not move at all. Approved access to a named app, and a standard you never have to see.

One thing to do before the next post.

Open the connectors list and count what you have already approved. If any of them is not serving a job you can name in one sentence, disconnect it. Then connect exactly one, for the thing you did by hand three times last week.

If you want the whole system rather than one piece of it, that is what I wrote Claude AI Bible (7 Books in 1) for. Seven books in one volume, with connectors and MCP in plain English, ten automation templates written in the read-and-prepare shape, and more than five hundred numbered prompts to copy.

This post is part of my Claude series. The one that started it is Why Claude Gave You a Great Answer on Monday and a Flat One on Tuesday.

Tomasz