Trust infrastructure for agents
Confirm a counterparty is eligible before your agent engages.
Elidian is a machine-consumable directory of verified counterparties. Your agents receive only those your human governors have pre-authorised — checked for credentials, standing and jurisdiction before any transaction begins.
Watch the film
Why verification before a transaction matters.
Average cost of one BEC attack
Organisations hit by BEC
Manual vendor onboarding
Staff time per vendor
How it works
From eligibility rules to an audited shortlist.
Set the eligibility rules
Your governors configure eligibility, policy-compatibility, authority limits and value thresholds for each querying agent.
Query by capability
An AI colleague submits the capability a sourcing task requires, and the directory returns matching candidates from its registered listings.
Filter to the cleared
Candidates outside your rules and authority limits are removed, leaving a shortlist your governors have already pre-authorised.
Approve new listings
New counterparties, and cases the checks cannot resolve cleanly, escalate to a named human before they are treated as trustworthy.
Keep the record
Every query, result and selection is logged, so a governance or audit review can reconstruct who was eligible and why.
Why Elidian
What sets the eligibility layer apart.
Stops at the gate
Elidian verifies standing before a transaction and goes no further. It does not negotiate, contract or settle, and holds no funds.
Named humans stay accountable
Every eligibility rule, onboarding approval and disputed verification traces back to an accountable person who owns the fiduciary duty.
Useful before scale
Adoption does not require a large ecosystem. Even a modest set of verified listings already serves real discovery needs.
Audit-ready by default
Discovery and selection events are recorded as they happen, giving procurement teams a retrievable trail rather than an after-the-fact reconstruction.
Consumes upstream identity
It builds on verified identity and credential assertions from the layer beneath it, rather than minting its own identity claims.
Inside the directory
The components behind a verified shortlist.
Verified Listings Registry
Structured listings of registered companies and their agent capabilities, each bound to verified credentials, jurisdictional metadata and machine-readable policy attributes.
Capability Matching Engine
Matches an incoming query to registered listings against a normalised capability taxonomy, so candidates stay comparable across different companies.
Credential Verification Service
Checks a listing against submitted and third-party credential sources, drawing on upstream identity assertions rather than issuing its own.
Eligibility & Policy Filter
Applies governor-configured rules, policy-compatibility attributes, authority limits and value thresholds so an agent sees only counterparties it may consider.
Trust & Reputation Records
Builds reputation from logged interaction outcomes, so trust standing reflects behaviour over time rather than a one-off check.
Staleness & Consistency Scanner
Continuously scans for stale, expired or inconsistent listings and flags or suppresses them before an agent treats one as eligible.
Onboarding & Approval Workflow
Routes new counterparties, and ambiguous or high-stakes verification cases, to human approval rather than auto-approving them.
Audit & Governance Log
Records discovery and selection events to support procurement audit, dispute resolution and governance review.
Discovery is the gate, not the deal
Every inter-agent relationship begins with a question an agent cannot yet answer for itself: is this counterparty legitimate, appropriately credentialed and cleared for my venture to engage? Today the candid answer is that discovery is ad hoc. An agent can locate a plausible provider, but it cannot readily confirm that provider's standing, so the work falls back to human-led, questionnaire-driven onboarding — or it does not happen at all.
Elidian answers that first question in a form an agent can consume, and stops there. It does not negotiate, contract or settle; those functions sit in adjacent layers. The boundary is deliberate, because a directory that verifies standing is useful on its own, before the wider ecosystem is large, and it becomes more valuable as more parties list and query.
Accountability stays with named people
AI colleagues run the operational work: verifying listings against submitted and third-party sources, matching queries to capabilities, scanning for listings that have gone stale, and maintaining reputation from logged outcomes. The decisions that carry fiduciary weight remain with people. A governor sets the eligibility rules, approves who is onboarded, adjudicates a disputed verification, and owns what the directory certifies as trustworthy. Ambiguous or high-stakes cases escalate to human review rather than being auto-approved — the intent is to place judgement exactly where the checks cannot resolve a case cleanly.
What we are still testing
We prefer to state the evidence as it is. The closest measured proxy for the harm we address is business email compromise, which the FBI's IC3 reporting puts at roughly USD 2.9bn in reported US losses in 2023 and about USD 2.8bn in 2024. That is human email fraud, not agent-to-agent risk, so it frames the cost of an unverified counterparty rather than proving the exact case. Whether that pain transfers cleanly into agent commerce, what buyers will pay, and how capable incumbents respond are open questions we are actively testing — not claims we ask you to take on trust.
What serious buyers want to know.
Start a conversation about agent-eligible sourcing.
Tell us how your agents source counterparties today, and we will walk through what a verified, pre-authorised shortlist would return.