ASK TOBY

Your agents. Your machine.

You can run an AI agent crew on a machine you own.

Your files stay with you. Your keys stay with you. Your logs stay with you.

This guide is a setup you host yourself, in the order that works.

1. Decide what must stay home

Start with a list, not a machine.

Write down what you would not hand to anyone else.

Client files. API keys. The record of what your agents did and why.

Those three go on your hardware. Everything else is a choice.

Your list is probably short. That is good news. A short list means a small machine is enough.

2. Start with one small machine that stays on

You do not need a rack.

One small machine, plugged in, with a stable connection, runs a crew.

Agents mostly wait. They read, call a model, write a file, and wait again.

That is light work for a desk sized box.

Pick one you can leave alone for weeks. Quiet, cool, and boring is the goal.

3. Split the work: your box or a hosted model

Two jobs sound alike and are not.

Running the agent is one job. Running the model is another.

The agent, its files and its logs live on your machine.

The model can be a hosted one you call, or one you run yourself.

Decide per task. Sensitive text stays on a model you control. General drafting can call out.

Write the rule once so every agent follows the same line.

4. Keep the keys on the box

Put every key in one place on the machine.

Never in a prompt. Never in a note an agent can read back. Never in a shared folder.

Give each agent only the keys its job needs.

If one agent goes wrong, the damage stops at its keys.

One place also means one place to rotate when you need to.

5. Backups before agents

Set up the backup before the first agent runs.

Then do the part most people skip: restore it.

Here is the test.

Restore a backup to a scratch folder. Open the file. Check it is the one you expect. Delete the scratch copy.

A backup you have never restored is a hope, not a backup.

Repeat the test on a schedule you will actually keep, and write the date down.

6. Reach it from your phone, and only you

You will want to check on your crew away from your desk.

The pattern is simple. The box is reachable only by you.

Nobody else gets a way in, and the machine does not advertise itself.

Decide that first. Then choose how to do it, and keep that choice out of your notes and your posts.

If you cannot say who can reach the box, it is not ready.

7. Add a second machine when the first is full

Do not buy for a future you have not met.

Run one box until it is the limit. Not a feeling. A real limit you can name.

When it is full, add a second box for one clear job.

Keep the same rules: your keys, your logs, a tested restore.

Growth is easy when the rules were set on day one.

Next

See your whole crew on one screen

Decide what needs your call

Get the free field guide and set up your first approval list.

Your Crew, Your Call field guide cover
FREE · PDF

YOUR CREW, YOUR CALL: THE FIELD GUIDE

How I run a fleet of AI agents from one screen and my phone.

One email with the guide. No spam, no list.

Want the full playbook? The Blueprint has the templates I use.

Questions people ask

What is local AI?
It's AI that runs on hardware you own, in your house. Your questions stay on your network, and no company can switch it off. Cloud AI is renting. Local AI is owning the house. If it runs on your machine, it answers to you.
Do I need a server room to run AI agents at home?
No. A few machines on a shelf will do it. Put them on one private network, and they work like one shop. Start with the computer you already have, and add when you outgrow it. A shelf, not a server room.
Cloud or local: which one do I use?
Both, and I choose every time. Private things stay local. Heavy lifting can go to the cloud. The point isn't never using the cloud. It's that I get to decide. Rent when it makes sense. Own what matters.

Jeff Sperandeo. Video editor. I ran the paid ads side and the organic side of a big YouTube channel at a Fortune 500 company. I saw how ads and AI find people from the inside. Now I build agents on my own machines, and I show you how.