ANID

Six layers of identity from keypair to runtime divergence detection.

Six layers of identity from keypair to runtime divergence detection.

Identity

2026

3 Weeks

by

Ash

brown sand with water droplets

Identity Beyond the Keypair

A bare key, DID, or credential answers only one question: who holds the key. In a world where agents move money, that is not enough. Nien's ANID — Agentic Network Identity — answers four: who is accountable, what is actually running, what is it allowed to want, and is it still acting like itself right now.

The four layers that matter

L0 binds the agent's keypair through a signed ownership chain to an accountable legal entity. Every task the agent submits is signed — unsigned tasks are rejected before they are evaluated.

L1 attests what is actually running: a hash of the agent's model, system prompt, tool manifest, and code. Change the prompt — the agent's DNA — and the attestation changes. Re-attestation is required before the agent can act again.

L2 binds the agent to a signed intent manifest: declared purpose, permitted recipients, amount and time envelope. Authority is scoped to that manifest, never granted in blanket.

L3 is runtime governance — the layer existing systems lack entirely. Every action is scored for divergence from declared intent. A hijacked agent holding a perfectly valid key is caught the moment its behaviour drifts. This is the differentiator.

More PRIMITIVES

More PRIMITIVES