Public Homepage

AMOF 3.0 Runtime Authority is live.

AMOF now owns runtime truth across intake, context, runners, scans, bounded loops, receipts, and evidence - while cognition workers remain replaceable.

AI agents are cheap. Runtime truth is not.

AMOF is the governed orchestration layer that controls context, execution readiness, policy attribution, receipts, and evidence before cognition workers mutate anything.

AMOF v3.0.0BOUNDED_RUNTIME_STABLE_LOW_RISK_EXPOSURE_WITH_KNOWN_GAPS_AND_CLEAR_NEXT_SLICESScan/report and bounded-loop-first public surface

Current Runtime Truth

console-cloud-dev.amof.dev
LOW_RISK_EXPOSURE
Public-safe runtime surfaces are reachable with known bounded gaps tracked as follow-ups.
ial-cloud-dev.amof.dev
SAFE_PUBLIC_SURFACE
Auth-bound remote surface reference; not framed as an open playground.

Public scope does not claim production readiness or autonomous mutation/dispatch.

Runtime Authority

AMOF does not treat chat output as runtime truth.

Runtime truth is recorded through inspectable runtime surfaces. Cognition workers remain replaceable components inside governed runtime loops.

Context selection

Explicit runtime context selection is part of operator-visible runtime truth.

Governed intake

Messy work is normalized through validated intake before execution readiness decisions.

Runner registry

Runner metadata and capability matching are first-class runtime evidence surfaces.

Execution scan/report

Readiness scans and reports expose blockers without dispatching remote execution.

Bounded loops

Long-running loops are controlled by stop conditions, policy discipline, and evidence.

Run inspection

Runs are inspectable with runtime-attribution context rather than chat-only claims.

Runtime logs

Structured runtime logs remain part of the authority model and evidence workflow.

Receipts

Receipts preserve runtime facts, outcome boundaries, and policy-linked provenance.

Evidence surfaces

Public-safe evidence/debug surfaces are exposed without private topology or secrets.

Operator Console preview

Cloud-dev console preview shows current runtime surfaces without fake demo overlays.

Governed Intake

Messy work enters through contractized intake, not ad hoc prompting.

AMOF validates and normalizes intake packets, preserves planning-only behavior when required, and routes work toward governed execution readiness.

Context Discipline

Context is explicit and fail-closed.

When remote/cloud context is unavailable, AMOF fails closed with explicit error outcomes. No silent fallback to local execution.

Runner Registry

Workers are metadata-driven and replaceable.

Runner metadata, capabilities, and matching are runtime-owned surfaces. AMOF owns coordination authority; individual cognition workers do not.

Execution Scan / Report

Readiness insight without remote dispatch.

AMOF can preview readiness, surface blockers, and produce scan/report outputs while preserving non-execution truth markers such as NO_EXECUTION_PERFORMED.

Bounded Loops

Controlled long-running loops with stop conditions and evidence.

AMOF bounded loops are policy-disciplined and evidence-linked with NO_MUTATION_PERFORMED and NO_REMOTE_EXECUTION_DISPATCHED boundaries.

Receipts and Evidence

Receipts preserve runtime facts with public-safe metadata.

AMOF preserves cost truth and safety boundaries: unknown cost remains unknown/null (never fake 0.0), with no secrets, no raw prompts, and no raw provider generation IDs in public surfaces.

Operator Console Preview

Cloud-dev live preview

The cloud-dev Operator Console exposes AMOF runtime receipts, intake submissions, selected runs, policy attribution, and sanitized evidence/debug surfaces. It is a live preview over the current AMOF runtime path, not a fake demo surface.

Cloud-dev preview. Public-safe runtime surfaces only. Known gaps are tracked as follow-up slices.

Current Known Next Slices

Runtime logs viewer contract and minimal UI
Add contractized viewer coverage to align operator console runtime-inspection experience.
Receipt count semantics contract
Clarify list/count semantics so receipts are interpreted consistently in public-safe surfaces.
Console rollout guardrail
Compare deployed hash against intended source before rollout to reduce regression risk.