Deny is the default
We'd rather block a bad action than apologize for it. In the runtime and in how we ship: fail closed, then earn the open.
Careers
Build the layer that makes enterprise AI accountable.
Agents are getting capable faster than enterprises can trust them. We build the runtime that closes that gap — governance, provenance, and control as first-class engineering, not compliance theater.
Remote-friendly · engineering-led · governance is the product
How we work
The values are the architecture.
You can read our engineering culture straight off the product — the same principles gate our own work.
Provable beats plausible
Claims carry receipts — in the Trust Ledger, in our benchmarks, in our numbers. If we can't demonstrate it, we don't say it.
Ship the whole system
A feature isn't done until it's governed, observable, and documented. Small team, deep ownership, no throwaway demos.
Humans stay in command
We automate aggressively and keep judgment human — in the product's oversight modes and in how we make decisions.
Open roles
No open listings right now — the bar stays open.
We hire deliberately and post roles when they're real. If governed agentic AI is the problem you want to spend years on, don't wait for a listing: send us the thing you've built that you're proudest of and where you'd raise our bar.
Engineering · Design · Solutions · GTM — speculative applications welcome via the contact page.
Tell us where you'd raise the bar.
Two paragraphs: the hardest system you've built, and what you'd fix about ours after a week inside it.