You set the limits.
An agent is only useful if you can trust it.
Minimum permissions
It only accesses what you connect.
Answers with sources
It shows where every fact comes from.
A log of every step
What it did, when and why.
Human approval
Sending, deleting or paying goes through you.
Your data, in writing
We agree what is stored and for how long (GDPR).
It always says it's an AI
If it talks to a customer, it says so.
AUTONOMY
Three levels of freedom.
LEVEL 1
Read
Reads and summarizes. Touches nothing.
LEVEL 2
Propose
Prepares the work and waits for your approval.
LEVEL 3
Act alone
Acts only where you allow it, and tells you after.