Verify

Check it yourself

Obelisk publishes evidence, not adjectives. Every claim on this page is something you can verify from your own machine, without an account and without trusting us. Nothing here asks you to believe a number — each one hands you the method.

Last updated 2026-08-27

Five things you can check right now

Each of these is independent. You do not need an Obelisk account for any of them, and none of them require our cooperation at the moment you check.

WhatWhereWhat it proves
A receipt is in the ledgerInclusion widgetYour browser refolds the Merkle path with its own WebCrypto — we never tell it the answer.
The ledger has not been rewrittenTransparency headPin the head today, compare it tomorrow. A rewritten history cannot produce the same root.
A token really came from usToken verificationLocal ES256 verification against the published JWKS. No call back to Obelisk.
A project is a registered relying partySeal checkThe seal resolves to a registry entry, or it honestly does not.
An agent's claimsProof LinksIndependent, individually-scoped claims with freshness — never a single "this is not a human" verdict.

What these proofs deliberately do NOT claim

An honest proof states its own limits, so here are ours in plain terms. A receipt proves an action was recorded and has not been altered since — it does not prove the action was wise. A transparency head proves the log is append-only between two points in time you checked — it does not prove we never had the ability to rewrite it before you started pinning. An agent Proof Link proves specific, individually-scoped claims — key control, an accountable owner, workload evidence — and never that a counterparty is not a human, that autonomy is permanent, or that a system is free from compromise.

We publish these boundaries because a proof that overclaims is worse than no proof: it moves the trust you withdrew from the vendor onto a artifact that cannot carry it.

The honest operator boundary

Obelisk runs on infrastructure Obelisk controls. An operator with host access is inside the trust boundary of anything that is not cryptographically bound outside it — which is exactly why receipts chain to a fixed genesis, why the transparency head is published for third-party pinning, and why token verification happens locally against a public JWKS rather than by asking us. Read the operator boundary for what that does and does not cover.

Go deeper

The full security posture · How trust is established · Live service + release proof · Build against it