Your own AI harness,
one paste away.
Copy one prompt, paste it into Claude, and it builds you a personal command center: many threads at once, all your own Claude seats, your own API keys in a sealed vault, and your own tools connected.
Then paste it into Claude Code and press Enter. The first thing it asks:
Three steps. About as hard as ordering lunch.
Claude does the building. You make the choices and sign in to your own accounts.
Copy and paste
Press Copy prompt, open Claude Code (the Code tab in the Claude desktop app, or claude in a terminal), and paste it into a new conversation.
Name your harness
Call it anything you like. Claude uses your name for the app title, the folder, the launcher and everything else.
Connect your own accounts
Sign in your own Claude seats, add your own Groq and other keys, link your own tools and bots. Claude tests each one as it goes.
What your harness gets
Everything we learned building ours, written down so Claude can build it for you, one tested piece at a time.
Many threads at once
Start, name, pause and resume conversations, each with its own folder.
All your Claude seats
Connect several of your own logins, see usage, and move work when one runs out.
Sealed key vault
Groq, OpenAI, Gemini and more, encrypted and only handed to a job after you approve.
Project threads
A lead thread starts sub-threads, shares a task list, and gets their reports back.
Done means done
Every reply ends with a done line, and unfinished work is nudged back on track.
Question cards
When Claude needs a decision you get tap-to-answer choices with a recommended pick.
Memory
Lasting notes about you and your projects, read at the start of every thread.
Approvals inbox
Nothing spends, sends, deletes or publishes until you click yes.
Web and proof pictures
Reads the public web, and screenshots the pages it builds before calling them done.
Voice and video
Talk instead of typing, have replies read aloud, and let it watch video links.
Recurring jobs
Scheduled prompts, daily recaps, and a ping when a long job finishes.
Your other devices
Optional sync to your other computers and phone, on your own server.
Yours, not ours.
The prompt is just instructions. It holds no keys, no accounts and no one else's data, and it tells Claude to refuse any key that is not yours.
You bring your own Claude seats, your own API keys and your own connections. Nothing phones home to us.
- Your name for it, not ours
- Your own Claude seats, signed in by you
- Your own Groq and other keys, sealed
- Your own tools and bots connected
- Asks before spending, sending or deleting
Read it first if you like
This is the whole prompt. No hidden steps.
Questions
What do I need before I start?
Claude Code (included with paid Claude plans, inside the Claude desktop app or as a terminal command) and a computer you can install things on. API keys are optional; Claude helps you get only the ones you want.
Does it work in the regular Claude chat?
It needs Claude Code, because it creates files and runs commands on your computer. If you paste it into a plain chat, Claude will tell you to switch to Claude Code.
How long does it take?
The basics usually come together in one sitting. The upgrades are built one at a time, and a build plan file lets any new conversation pick up where the last one stopped.
Will anyone else see my keys or chats?
No. Everything runs on your own machine or your own server, your keys stay sealed in your own vault, and the prompt tells Claude never to show a key in chat.
Can my partner or team use it too?
Yes. Each person gets their own profile, their own seats and their own keys. Nothing is shared unless they add it.
Ready to name yours?
One click, one paste.