How we work
How we work: you decide where every agent stops.
Before anything runs, we set in writing for your project what an agent may do alone and what needs a person.
The path
From one page to a system that runs.
One page
From your website and public data. Free.
5-Day Data Check
Your real exports, under NDA. Findings in writing.
Scope in writing
What gets built, what does not, who approves what.
Build
Agents and automation on your data, step by step.
Run
We operate what we build and you can see what it did.
Where the agent stops
Five levels. You pick one per workflow.
- Read only. The agent reads and stops. A person sees what it looked at.
- Count only. The agent reads and counts, then stops. A person sees the list.
- Draft only. The agent prepares a draft and stops. Nothing leaves without a yes.
- Send alone. The agent sends by itself. Only for steps you released in writing.
- Run alone. The agent runs to the end. Only for steps you released in writing, and every action is logged with a way back.
Most first agents start at Draft only. Anything above that needs your release in writing.
Ownership
Who owns what.
This is what we set in writing for your project.
Audit trail
One real example.
Our own Google Ads tool runs on the Google Ads API with Basic Access, in Germany. It changes a campaign only after a person approves it, the budget lock sits in code, and every change is logged with a way back. See what runs today.
Limits
When this is not right for you.
- You want everything automated with nobody looking. We will not set it up that way.
- You need a guaranteed response or uptime figure. We do not promise numbers we cannot hold every day.
The rules are yours.
We start with one page about your company.