About the Registry

The Agentic Foresight Registry is a public registry of automated task risk assessment for autonomous agents. Every assessment is computed by a fixed algorithm named AFR-1, issued immediately as a permanent numbered certificate, and priced as plain text at every step of the process — Assessments from EUR 1.90.

This page states what the Registry issues, where the method that computes it comes from, what it does not claim, and how it is operated. Nothing below is a summary written after the fact; each section names the page where the underlying fact can be checked.

Read the method → View the registry →

3

Assessment tiers published

0

Visitor identifiers collected

0

Language models on the delivery path

1

Published method (AFR-1)

What the Registry issues

The Registry issues one kind of document: a numbered certificate recording a computed task risk assessment. Three tiers exist, each a one-off purchase with no subscription and no usage metering, distinguished only by how much of the same computation the certificate carries.

TierPriceContains
Standard Assessment EUR 1.90 Verdict, Recommended window
Extended Assessment EUR 4.90 Verdict, Recommended window, 24-hour outlook, Risk factors, Tool compatibility
Full Chart Assessment EUR 14.90 Verdict, Recommended window, 24-hour outlook, Risk factors, Tool compatibility, Permanent chart

Full machine-readable pricing is always available at /pricing.json, with no authentication required, and the same figures render for a human reader at /pricing.

The method's provenance

Every assessment on this site is computed by AFR-1, and AFR-1's full derivation is published rather than described.

Deterministic

AFR-1 is fixed arithmetic over published tables. There is no language model anywhere on the delivery path, and task classification is a plain select field for exactly this reason.

Discarded on receipt

The task description submitted with an order is reduced to a one-way hash. The original text is never written to the database, to logs, or to any error response.

Published in full

The complete derivation — every table, every step from submitted parameters to computed verdict — is documented in the open at /method, not summarised after the fact.

The complete pipeline, every reference table, and the reasoning behind each step are documented in full at /method.

The declared limit

The Registry does not test language models. It has never executed a task on one, and it holds no record of what happened when a task it assessed was actually run.

No assessment issued by the Registry is validated against a real-world outcome, and none carries a representation that it predicts one. This is not a qualification added here for the first time — it is stated in the same terms in the Service description at /terms, and the reproducibility statement at /method makes the same distinction from the other direction: the claim made for AFR-1 is reproducibility, not accuracy.

Operations

Six operating facts, each checkable at the page named beside it.

  • Operated from Finland. The Registry is sold by a private trader established in Finland. The statutory seller identity required by consumer-protection law is published in full at /terms.
  • EU data residency. Certificates and the engagement log are held in a database located in the European Union.
  • No visitor identifier of any kind. No IP address, no user agent, no cookie, no browser storage value, and no fingerprint is collected — of a purchaser or of anyone reading this page. Full detail is at /privacy.
  • A public engagement log. Every logged visit to the site is recorded as an event, never as a visitor, and the log itself is published at /engagements.
  • A public certificate registry. Every issued certificate's existence — its number, issue time, tier, model and task class — is published at /registry. Its computed content is not.
  • Certificate verification. Any issued certificate can be checked for standing and issuance at /verify using its printed verification code.

Prepared by The Agentic Foresight Registry

Frequently asked questions

Does The Agentic Foresight Registry call a language model to compute an assessment?
No. The delivery path is deterministic arithmetic over fixed tables, published in full at /method. Task classification is handled as a plain select field for exactly this reason.
Is the task description submitted with an order stored anywhere?
No. It is reduced to a one-way hash on receipt, and the original text is discarded immediately — it never appears in the database, in logs, or in an error response.
Does an assessment predict what will actually happen when the task runs?
No representation of that kind is made. The Registry has never executed a task on any model, and no assessment it issues is validated against a real-world outcome — stated in full at /terms.
What does the Registry collect about a visitor to this site?
No identifier of any kind: no IP address, no cookie, no browser storage, no fingerprint. The engagement log at /engagements records events, not visitors.

Get an assessment computed by this method

Three tiers, one method, no subscription. Assessments run from EUR 1.90 for a Standard Assessment to EUR 14.90 for a Full Chart Assessment that includes the permanent chart itself.

Get Assessment