Ujex is the trust layer for AI agents.

Proof, not promises.

Every action an AI agent takes is identified, gated by the approvals you require, and written to a tamper-evident, Ed25519-signed, hash-chained audit record you can verify offline. Open-source (Apache-2.0), with SDKs in Python, Go, and TypeScript: Passport, Approvals, Governor, Secrets, Metering, and a permanent record no one can quietly rewrite.

Trust primitives

Evidence outputs — signed AAT bundles, manifest signatures, anchors, and agent passports
Agent Inbox — per-agent email with risk, approval, and evidence links
Provenance Memory — markdown-first context with provenance labels
Budgets — estimated action-based spend and risk checks before agents act
Evidence — hash-chained log, hourly verified, exportable for auditors

Get started

Quickstart
Full docs
OpenAPI spec
Pricing
Evidence outputs
Security model
FAQ

Compare with

vs AgentMail
vs HumanLayer
Ujex alternatives by trust primitive

Enable JavaScript for the full interactive experience.