Developers

Two local tools. Two different questions about AI evidence.

AgentProof leaves a receipt of what a collector observed during an execution. EvalDossier checks what basis an evaluation dossier declares and whether its structure, signatures and bindings match.

This page lists what exists, where it can be installed and what must not be inferred. Directory states include dates so a pending review cannot silently look current forever.

Current projects

Choose by the question you need answered

Agent-session receipts

AgentProof

Creates local, hash-only receipts of collector-observed events and checked repository commitments, then compares those commitments with the current repository.

Does not establish

It does not establish authorship, exclusive causation, completeness, correctness, compliance or producer-independent tamper evidence.

Evaluation-dossier verification

EvalDossier

Locally checks the structure, signatures, bindings and declared basis of compatible evaluation dossiers. It preserves INCONCLUSIVE as distinct from established false.

Does not establish

It does not determine whether the work is good, the evaluator is honest, a claim is true or a payment should be released.

Availability checked 28 Jul 2026

Status by product and host

“Complete” is not presented as “published”. If no public installation link exists, this table says so.

ProductHostStatusBasis
AgentProofCodexNot listedCandidate complete; no public install link.
AgentProofClaude CodeReview pendingSubmitted to Anthropic on 27 Jul 2026; source public.
AgentProofOpenClawAvailableClawHub v0.1.0.
EvalDossierCodexNot listedCandidate v0.2.1; no marketplace link.
EvalDossierClaude CodeReview pendingSubmitted to Anthropic on 26 Jul 2026; candidate v0.2.2.
EvalDossierOpenClawAvailableClawHub v0.2.2.
Historical HREVN runtime

Narrow public surfaces, described as narrow

These URLs still respond, but they are not the complete AgentProof or EvalDossier contract. No supported public MCP server is advertised until a documented, usable surface exists.

Health endpoint

Confirms only that the historical HREVN runtime is reachable.

Open health JSON →

Discovery OpenAPI

The current public specification documents the health endpoint. It is not presented as a complete product contract.

Open openapi.json →

Historical manifest

A discovery document for the HREVN runtime. It is not an AgentProof or EvalDossier plugin.

Open manifest →
Technical FAQ

Questions that change how a result should be interpreted

Are AgentProof and EvalDossier the same product?

No. AgentProof records commitments about an observed execution. EvalDossier checks the form, integrity and declared basis of a compatible evaluation dossier.

Do all three hosts produce identical evidence?

No. They share a format and discipline, but each collector declares its own capabilities and limits. Codex and Claude Code start a new session; OpenClaw observes the current run through hooks.

Are these tools an audit or legal evidence?

No. They are local technical tools with narrow claims and explicit limits. They do not replace investigation, independent audit, legal review or professional judgement.

Does the public API expose every HREVN function today?

No. The current public OpenAPI is a discovery surface that documents health. This page does not claim a complete public API or MCP contract.