Products
What we make
Vardhold
Endpoint governance for AI agents
Sees every AI agent on a Windows machine, stops dangerous actions, holds risky ones for a person to approve, and keeps tamper-evident proof in your SIEM. Works with Claude Code, Cursor, Copilot and MCP servers.
Explore Vardhold →More in the lab
Vardhold is our first product. We are working on more tools built the same way. Tell us what you need, and you may help shape them.
Get in touch →How we build
Principles that hold in every product
Your data stays yours
Our software runs on your machines. It sends nothing to us, and what it records goes to systems you control.
Secure by default
When something is unclear, the safe choice wins. Protection is on from the first install, and loosening it takes a deliberate step.
A person decides
Automation handles the routine. Anything with real consequences waits for someone who can be held to account.
Honest about limits
We say exactly what our products prevent, what they only detect, and what they do not cover yet.
Contact