Using the product

Governance screens

The Gate records real PIE fix-gate verdicts as signed decisions; Trust Boundaries treats your settings as policy, with mutations routed through a separate signed flow.

The Governance group is where you prove that the bad thing was prevented, and where your policy lives as policy rather than as toggles.

The Gate

/the-gate · reads GET /api/gate/events

Every row on The Gate is a real remediation the Policy Intelligence Engine evaluated — this is where you prove "the bad thing was prevented." Each verdict is anchored as a signed gate-decision ledger entry. The four decisions are:

  • Allowed — merged, PIE allowed it.
  • Mediated — allowed, but gated on a human merge.
  • Denied — blocked, rejected, or failed.
  • Pending — no verdict yet.

A metric strip summarizes gated fixes, allowed/mediated/denied/pending, and the last 24 hours.

This screen is the fix-decision stream — a PIE verdict on each of your remediations, not a live runtime traffic proxy. Per-request runtime enforcement (the optional inline Gate proxy that allows or denies live identity and agent requests in real time) is a separate runtime tier; deploy it and live requests stream here alongside these fix verdicts. There is no fabricated live ticker on this screen.

Trust Boundaries

/trust-boundaries · reads GET /api/trust-boundaries

Trust Boundaries treats your settings as policy, not toggles, in three sections:

  • Policies — the rules The Gate evaluates, each with an effect (ALLOW / DENY / MEDIATE), scope, pinned status, and enabled state. The "evals · blocks / 7d" column is runtime-tier telemetry reported by the Gate.
  • TTL & rotation — the default credential TTL per credential kind.
  • Hard guardrails — org-level, locked guardrails that are admin-gated and ledger-anchored.

Mutating controls here (toggles, + New policy, sliders) are intentionally disabled on this screen — real changes go through a separate, signed flow so a policy change is itself a provable, ledger-anchored event. An org with nothing configured shows an honest "No trust boundaries configured yet" state.