NautLoop
Sovereign agents in sandboxed execution. Your data never leaves the boundary you drew. Swiss-hosted, privacy by architecture rather than by promise.
01 / The build
The boundary is the product.
Agents run inside a sandbox with no outbound route you did not open. The model runs where you put it, the audit log is written before the response is returned, and the whole thing is hosted in Switzerland.
Built for the room where someone from risk asks where the code goes.
Why I built it
Every agent framework assumes your code goes to someone else's GPU. For a regulated business that assumption is the entire conversation, and it ends it. NautLoop starts from the opposite default.
02 / Design decisions
The trade-offs are part of the story.
01Deny egress by default
The sandbox opens nothing unless you say so. It makes first-run harder and it is the only version of the promise that survives an audit.
02Audit before response
The log is written before the answer is returned, not after. Slightly slower, and it means a crash cannot lose the record of what happened.
03Not open source, and said plainly
There is no public repository for the core. Calling it open source because adjacent repos are public would be the kind of loose claim this whole site argues against.