About
A research lab for verifiable private computation.
Urizen Research Labs builds machines that compute on encrypted data and prove their work — so the result can be trusted without the data, the logic, or the operator being exposed.
What we work on
The lab advances verifiable private computation: machine-checked theory, confidential execution, and proof-carrying systems. From that research we build Nibiru, an engine for private execution with public proof of correctness; Baobab, the verifiable control layer for AI agents; and Anansi, the harness that joins them into governed agentic AI. Heimdall is our research demonstration of privacy-preserving identity checks built on the same foundations.
How we work
Theory first: the proof theory behind the systems is machine-checked in Lean 4, end to end. Builds are reproducible and signed, so a release can be verified byte for byte. Claims on this site carry exact status labels — Proved, Verified, Implemented, Specified, Conjectural, Open, Unresolved — and a demonstration is always labelled as what it is. The Nibiru demo build cleared 100 consecutive transactions over a continuous 24-hour endurance run.
Where we are
The systems are experimental research previews and are not yet generally available. Nibiru runs its full custody-to-settlement transaction today as an experimental MVP. Baobab is in early access with design partners. The Heimdall demonstration is not a regulated KYC product or production service. If you want to build with us, request access.
Find your way
Start with the research overview, see the interactive Nibiru demo and the Anansi harness demo, or read the engine notes and docs. Developers and AI agents: the machine surfaces are indexed at /developers/.