Skip to content

Site search

Type to search Pages

Reliance for AI agents

Your agent can rely on a counterparty's claim — without re-verifying it.

Signura is a neutral exchange where issuers sign attestations they stand behind, and your agents verify the proof at the point of decision, retain it as evidence, and log every reliance for your governors.

Hero illustration for the home page of Signura

Watch the film

Five parts, one accountable flow.

Issuer attestation service

Abstract product-interface illustration for the app attestations page of Signura

An issuer creates signed attestations against verified source facts, sets each one's scope and liability terms, and revokes it when a fact changes. AI colleagues run this against issuer-approved policy.

Reliance and verification API

Product-interface illustration for the app dashboard page of Signura

A relying agent queries the exchange, verifies a proof cryptographically, and proceeds or escalates according to policy-as-code. Charged per reliance event.

Freshness and revocation engine

Product view (demo data) — nav-escalations in Signura

Revocation and freshness semantics — extending the proven OCSP status pattern to assertions — let a relying party tell a currently valid claim from a stale one.

Evidence and audit ledger

Product-interface illustration for the app evidence page of Signura

Every reliance event is written to a tamper-evident trail showing what evidence underpinned each decision, ready for governance, liability and dispute resolution.

How it works

From a signed fact to a logged decision.

Issue and publish

An issuer's governor approves the facts and liability the organisation will stand behind. An AI colleague then signs attestations against verified source facts, sets scope and expiry, and publishes them for relying parties to consume.

Rely at decision time

When a counterparty claim matters, the relying agent fetches the proof, checks its freshness and revocation status, and either acts within its governor's configured limits or refers the case to a human.

Revoke and propagate

When an underlying fact changes, the issuer revokes the attestation and the engine propagates the change, so no agent keeps acting on a claim the issuer no longer stands behind.

Assemble evidence

For any past decision, a governor retrieves the reliance record to show exactly which attestations were verified — satisfying compliance, liability and dispute-resolution needs.

What sets it apart

Built to be relied on, and audited.

Neutral by design

The exchange is a shared dependency, not a competitor to either side. Issuers and relying agents connect to the same neutral layer, and it takes neither side's part.

A named party behind each claim

Transferable reliance without an accountable party is the gap Signura exists to close. Every attestation names who stands behind the fact and on what terms.

Priced against re-verification

Reliance costs a fraction of a fresh verification, because the checking work has already been done and signed for.

Standards, not a private protocol

Built on the W3C Verifiable Credentials 2.0 substrate, a formal Recommendation ratified in May 2025, and aligned with the direction of eIDAS 2.0.

Accountability stays with people

AI colleagues run the recurring issuing, checking, propagating and logging; humans decide what to attest to and which issuers their agents may depend on.

Before you commit

Questions serious buyers ask.

Which inter-agent protocol do I have to commit to first?

Signura is a neutral layer built on the W3C Verifiable Credentials 2.0 standard. It sits above identity and credentialing primitives and beneath your commerce and settlement layers, so you can adopt it alongside whichever inter-agent protocol your agents use, rather than betting on one.

How do I tell a currently valid claim from a stale one?

Each attestation carries freshness and revocation semantics. Before it acts, a relying agent checks revocation and freshness status in real time, so it distinguishes a claim that holds right now from one that no longer does — not one that was merely valid when last cached.

Does a reliance layer just move liability around instead of removing it?

Each attestation carries explicit, scoped liability terms and a named party who stands behind the claim. Issuers decide what they will attest to and accept liability for before any reliance occurs, so accountability is defined up front rather than shifted silently — and every attestation states the limits of what was promised.

What happens when a reliance request falls outside its configured limits?

Governors set admissible issuers, attestation types, value thresholds and expiry. When a request falls outside scope, exceeds a value threshold, or has expired, the agent does not proceed on discretion; it routes the case to a human governor for review.

Is the exchange operator itself neutral?

Yes. Signura sits between issuers and relying agents and takes neither side's part. The operator does not become a relied-upon party beyond the integrity of the proof and the accuracy of freshness and revocation status.

Can an agent's action be attributed to a verified fact for audit?

The evidence ledger records which verified attestation underpinned each action — the claim, its scope, the liable party and the proof verified. An action is attributable to a specific verified attestation and its issuer, rather than to a shared credential.

What facts can issuers attest to at the start?

The initial focus is a narrow set of high-frequency common facts — licence validity, policy status, certification, and granted permissions — where verifying a fact once and relying on it many times removes redundant re-verification.

Give your agents something your governors can audit.

Tell us which counterparty claims your agents keep re-verifying, and we will walk you through how reliance would work on Signura.