1Pick a scenario
What this is: five industry actions, each fed to Synchronicity as raw facts only — no answer is embedded in the input. Why it matters: the verdict you see is derived by the policy, not echoed back from a field you sent. Pick one to begin — your first governed decision is one click away, no setup.
2The neutral facts you're sending (the SPAD)
What this is: the Structured Proposed-Action Descriptor — the inputs the caller transmits.
Why it matters: notice there is no ok, hold_reason, or verdict field here. Just facts
(e.g. loan_amount, account_age_days, fraud_score). The policy decides what they mean.
3Policy Lab — your rulebook, not ours
What this is: the licensee-authored policy this tenant is governed by — visible and editable. Why it matters: move the one highlighted threshold and watch the verdict flip in real time. That proves the policy is doing real evaluation and that KDH ships no built-in opinions. (This panel is a dry-run — it never signs or records.)
4Get the governed decision (the signed DDA)
What this is: the real /v2/decide call against the deployed, versioned policy, returning a
signed decision artifact. Why it matters: you get the verdict, a human-readable reasoning trace, the
policy-snapshot binding, the per-call latency, and a cryptographic signature — and Synchronicity took no action.
5Verify it yourself — offline, without trusting KDH
What this is: the trust anchor. The DDA is Ed25519-signed; the public key is published at
/v2/keys. Why it matters: your browser verifies the signature locally (no call to KDH), and replay
re-derives the same outcome against the captured policy snapshot — the auditor's "prove it two years from now" answer.
Verify from your own terminal (no service dependency)
Download the DDA, then run the bundled offline CLI against the public JWKS:
6Fail-closed — the unhappy paths
What this is: what happens when policy is silent or the service is unreachable. Why it matters: a control boundary that fails open is not a control. Synchronicity holds when no rule covers an action, and the calling pattern blocks on timeout/error — it never bypasses.
7Multi-tenant isolation & the audit export
What this is: each tenant is governed by its own policy and its own audit chain. Why it matters: a compliance officer can pull a tenant-scoped evidence file over a time window — the "hand the regulator a file" deliverable — and one tenant can never see another's records. The same signed evidence renders as a plain-English audit report (what an auditor actually reads) and as machine-readable JSON/CSV (what tooling ingests and what proves nothing was altered).
!What Synchronicity does not do
Stating the limits is part of being trustworthy. Synchronicity is a decision authority — nothing more.
Honest production signals: this is a real service running in the sandbox, CI-green, tested, and fail-closed. KDH is the architecture licensor, not the regulated party — Synchronicity produces the evidence and controls that help a licensee meet its own SOC 2 / HIPAA / GDPR / EU AI Act obligations in the licensee's own environment.
Run this in your environment
Don't watch ours — run yours. Drop the container in, wire the reference adapter in ≤ 20 lines, point it at a policy template, and get a governed decision on your own machine. Then let's talk.
Advanced: point the demo at a different Synchronicity host
Leave blank to use this same host (recommended). No trailing slash.