Platform · 07
Verifiers and outside access
How a company lets an auditor, certifier, buyer or authority read the evidence of chosen lots, for a limited time, and how anyone can check a lot or a document without an account.
What anyone can check
Some checks need no account and no permission from anyone:
- A lot's public record. Each lot has a public record without personal data: its product, quantity, dates, lineage, status and certificates with their evidence levels.
- A lot's information. Each version of a lot's information has a fingerprint on the ledger; anyone can recompute it from the published information and compare.
- A sealed document. The verification page of the platform says whether a report generated by A-REX Platform is unaltered and who sealed it.
Because the records follow the open standard, these checks can also be made with public tools, without relying on A-REX.
Verifier access
For more than the public record, a company gives a verifier access: read access to the evidence of chosen lots, for a regulator, an auditor, a certifier, a buyer or another verifier, often through the verifier's own software or AI agent.
The owner or an administrator creates it with a second factor and decides:
| Setting | Choices |
|---|---|
| Who and why | the verifier's name, kind and purpose |
| Which lots | chosen lots, products or a period of lot activity |
| What it shows | minimal: the lot's history, certificates, evidence levels and origin references; EUDR: adds the due diligence fields and the plot geolocation the regulation asks for |
| For how long | at most one year |
The verifier receives a credential once. It reads only the lots and the evidence the access covers, and nothing else: no directory, no personal data, no other company's records, no way to propose a change.
What the company sees
Every read by a verifier is recorded: what was read, when and from where. The access page lists them, and the company keeps the log for one year by default. Revoking the access stops the credential at once.
The same log covers every read by software outside the console: the company's own integrations, outside AI agents connected to its records and machine clients. Only the company's own team in the console and its assistant are not logged there; their actions go to the audit log.
Agents that verify
Verifiers increasingly send an agent rather than a person. A verifier's agent connects to the platform's MCP server with its credential and sees only two tools: list the lots it may read, and read a lot's evidence. It works with the same rules as every agent on the platform: records are data, never instructions, and evidence is never a verdict. See Developers and agents.
Further reading
In the help: Check a lot, Evidence levels.