Endpoints and integration paths.

All 12 endpoints in one file.

Lists every domain slug, version, control count, attestation artifact, SHA-256 checksum, and canonical endpoint URL. All 12 domains are at schema v1.1.0 with extended testability fields: validation_objective, evidence_required, machine_tests, human_review, blocking_effect, normative_status, anti_patterns, and update_status.

↓ manifest.json

All normative sources with canonical metadata.

Machine-readable registry of every standard, regulation, framework, and guidance document referenced in Apeiris control matrices — 335 sources including the OWASP AI Testing Guide v1 — with source_id, publisher, version, canonical URL, effective date, jurisdiction, normative force, and update cadence.

↓ source_registry.json

Every control that cites a given source.

The reverse of each control's crosswalk: pick any standard, regulation, or framework and see every control across all 12 domains that references it — citation transparency for verifying coverage.

→ Browse source references

Control proposals Apeiris submits to other standards bodies.

A dated, machine-readable record of controls and requirements proposed upstream to external control libraries and certification standards, each traced to its canonical apeiris:// source control with a neutral per-proposal status. Entries appear only after submission; the record is provenance, not a scoreboard.

→ Browse contributions  ·  ↓ submissions.json

The real taxonomy of every cited framework.

One machine-readable file per framework capturing its actual structure — requirement IDs, section titles, edition, and capture provenance — taken only from primary texts. Every anchored framework citation is validated against these anchors in CI, so a fabricated section number fails the build instead of surviving until an audit.

/integration/anchors/[framework].json

All 191 requirements mapped to Apeiris controls.

Maps every OWASP AI Security Verification Standard 1.0 requirement (C01–C12) to Apeiris control URIs with a coverage verdict: direct, partial, none, or out_of_scope.

↓ aisvs_coverage_matrix.json

RFC 9396 cross-domain validation profile.

Defines the 8 capabilities required for per-request structured agent authorization — registered identity, ephemeral credentials, permission ceilings, resource server validation, fail-closed enforcement, and policy-as-code governance. Normative anchor: RFC 9396 (OAuth 2.0 Rich Authorization Requests).

↓ profiles/structured_agent_authorization.json

Per-requirement coverage for major frameworks.

Explicit coverage manifests mapping every requirement of each major framework to Apeiris controls with direct / partial / none / out_of_scope verdicts. Available for: EU AI Act, ISO/IEC 42001, NIST AI RMF + AI 600-1, SR 26-2, DORA.

/integration/coverage/[framework]_coverage.json

Cryptographic integrity verification spec.

Specifies the algorithm (Ed25519), canonicalization method (RFC 8785 JCS), and verification procedure for integrity assertions on Apeiris domain control matrices. Per-domain SHA-256 content hashes in manifest.json are recomputable by any consumer over HTTPS, and the manifest is signed with an offline Ed25519 key — the public key is published and the signature is verifiable in your browser.

→ Verify integrity live  ·  ↓ profiles/manifest_signature.json

Six identity modes from session to attested workload.

Defines the six identity modes for AI agents — user session, brokered user delegation, service account, governed service account, attested workload identity, and hybrid user-to-workload chain — each with appropriate use cases, prohibited patterns, required Apeiris controls, failure modes, minimum evidence, and migration path. Includes SaaS authorization ceiling guidance, identity taxonomy (workload / personal / personal-in-enterprise), and provider implementation signals for Okta, Entra, AWS Bedrock, Google Agent Identity, SPIFFE, and Canva's Credential Broker pattern.

↓ profiles/agent_identity_mode.json

Evaluate an action against the full evidence fabric.

The platform layer (currently in design) will expose an attestation query interface — a single endpoint that accepts an action context and returns an evidence DAG verdict across all twelve domains.