How it works
How OverPowered works
One orchestrator runs the agents on all your machines. It works out the next step by analysing the work, then writes the prompt, the way you would. Here is that loop, step by step.
The loop, step by step
Your agents work on what you set
Your agents run on each of your machines, in your own AI agent tools.
It decides the way you would
It takes the decisions and writes the prompts in your place, the way you would, after analysing the work.
It knows what every agent is doing
The orchestrator knows what every agent on every machine is doing, so you don't have to check each one.
Critical decisions wait for you
By default, critical decisions such as payments wait for your approval. You can choose to give it full authority, payments included.
You follow it in the dashboard
See what every agent and the orchestrator did, and answer any critical decision that waits for you.
The work keeps going while you're away
The orchestrator keeps deciding and writing the next prompt, so your agents keep working on what you set.
From step 6, the loop goes back to step 2.
One orchestrator, every machine
One orchestrator runs the agents on all your machines, not one per machine or per session.
Go deeper
Agent org chart
Who does what: you, the orchestrator and the agents on your machines.
Governance and approvals
What it decides on its own, and what waits for your approval.
Set it up on GitHub
Go to the repository on GitHub and read its README.