ZeroPoint gives you cryptographically verifiable accountability — signed receipts, narrowing capability grants, and non-removable constitutional rules — without a central authority.
AI agents act across organizational boundaries at machine speed, yet the systems beneath them still rely on mutable logs, informal permissions, and platform-controlled identity. When something goes wrong, there is no shared, independently verifiable record of what happened or who authorized it.
ZeroPoint makes every action provable using three cryptographic primitives: hash-chained receipts, portable capability grants that strictly narrow with each delegation, and two constitutional rules that evaluate before every action and cannot be removed. These guarantees work identically across any transport.
Every action produces a signed, hash-chained receipt. The pr field references the predecessor, making the chain tamper-evident: modifying any entry requires rewriting every subsequent one and convincing every independent verifier to accept the rewrite. Chains verify offline, cold, with no running server.
Signed authorization tokens specifying scope, time window, cost ceiling, delegation depth, and minimum trust tier. Any grant holder can delegate a subset. Eight invariants enforce that authority narrows at every step — never widens. A revoked parent invalidates all descendants.
HarmPrincipleRule and SovereigntyRule evaluate before every action at every node. Fixed at positions 1 and 2 in the PolicyEngine — they cannot be removed, reordered, or overridden by any capability grant, WASM module, or operator action.
ZeroPoint is written in Rust and designed to be embedded into agents, services, and infrastructure. It ships with a Reticulum-compatible mesh transport, TCP/UDP interfaces, an HTTP API, and a dual-backend peer discovery layer called the Presence Plane.
Run a governed agent simulation and inspect live receipt chains in your browser.
Open the Lab →Integration guides, governance hooks, and practical code examples.
Read the Docs →Complete reference covering receipt format, delegation invariants, PolicyEngine design, Presence Plane, external truth anchoring, fleet topology, and threat model. Written for engineers.
Read Whitepaper v3.0 →Also available as PDF · ZeroPoint Open Foundation · June 2026