Decision plane · Ledger

Decision Ledger Explorer

Every decision our systems make is committed as a hash-chained ledger entry, so any output can be replayed, audited and defended. Search a synthetic ledger, trace a decision's lineage, replay it — then edit one digit and watch the chain refuse it.

Open the explorer Governance model

Synthetic ledger · seed 42

Decision records

240 synthetic records across fintech, healthcare, energy and federal workloads. The data is generated; the cryptography is not — every hash below is a real SHA-256, recomputed in your browser.

Chain intact240 recordstip 8,442,109

Synthetic demo data

Domain

Outcome

Guardrails

Reviewer

240 records

Synthetic decision ledger, 240 records, sorted by height descending. Select a decision ID to open lineage and replay.
			Context hash					Ledger hash	Parent
2026-09-24 16:01:43Z#8,442,109	foia-redaction-agentfederal	xgb-risk-v12	9597…855c	2 · pass	executed	0.924	a.okafor	faae…9583
2026-09-24 16:00:55Z#8,442,108	fraud-triage-agentfintech	bedrock/llama-4-70b-ft	d678…d1e4	3 · pass	escalated	0.624	j.lindqvist	0f89…aa6c	root
2026-09-24 16:00:00Z#8,442,107	meter-anomaly-agentenergy	claude-opus-5-5	f3bb…8fd2	4 · blocked	rejected	0.831	m.reyes	d48f…e927	root
2026-09-24 15:59:04Z#8,442,106	credit-limit-agentfintech	claude-sonnet-5	2d7b…ac88	3 · pass	escalated	0.616	j.lindqvist	4678…f567	root
2026-09-24 15:58:02Z#8,442,105	contract-clause-agentfederal	bedrock/llama-4-70b-ft	fd99…2d70	4 · blocked	rejected	0.581	m.reyes	a1b4…4502
2026-09-24 15:57:49Z#8,442,104	foia-redaction-agenthealthcare	bedrock/llama-4-70b-ft	81a3…b078	3 · pass	executed	0.769	—	d359…9677
2026-09-24 15:56:21Z#8,442,103	dispatch-optimizerenergy	claude-sonnet-5	baa8…f415	2 · pass	routed	0.884	—	5fb7…f6ed	root
2026-09-24 15:55:56Z#8,442,102	credit-limit-agentfintech	bedrock/llama-4-70b-ft	34be…a4f3	4 · pass	escalated	0.622	r.haddad	0c73…55ff
2026-09-24 15:55:46Z#8,442,101	credit-limit-agentfintech	bedrock/llama-4-70b-ft	4004…1fa5	2 · pass	approved	0.827	—	aeba…c62b
2026-09-24 15:55:32Z#8,442,100	fraud-triage-agentfintech	claude-opus-5-5	db76…e2da	3 · pass	escalated	0.595	r.haddad	9ed1…93ea	root
2026-09-24 15:55:08Z#8,442,099	prior-auth-agenthealthcare	azure/gpt-5-mini	7dd8…6a4d	3 · blocked	rejected	0.817	r.haddad	1d82…b3ad	root
2026-09-24 15:54:28Z#8,442,098	intake-summarizerhealthcare	claude-haiku-4-5	cfb3…06b5	2 · pass	approved	0.896	—	4df6…8824	root

Mechanism

Why the ledger is tamper-evident

Three properties do the work: canonical serialization, hash chaining, and replay from content-addressed inputs.

01 · Chain

Each hash commits to all history

ledgerHash = SHA-256(
  prevHash ‖ ":" ‖
  canonicalJson(record)
)

The record body — including prevHash — is serialized with sorted keys, so the same record always produces the same bytes. Because every entry folds in its predecessor's hash, changing any historical field changes every hash after it.

02 · Replay

Determinism you can re-run

Replay loads the context bundle by its hash, re-applies the recorded guardrail verdicts against their pinned policy versions, re-canonicalizes the body and recomputes SHA-256. If the result equals the stored ledger hash, the record is byte-for-byte what was committed at decision time. Open any record above and press Replay decision.

03 · Verification

Edits can't hide

verifyChain walks every record, checking each prevHash link and recomputing each hash. An edited record fails at its own height; recomputing its hash to cover the edit moves the failure to the next record, whose prevHash still commits to the original. Use the Tamper toggle to see it.

What each column means

Decision ID: Stable identifier for one committed decision. Deep-linkable: /ledger?decision=dcn_…
Timestamp: UTC commit time, with the record's height (monotonic position) in the chain beneath it.
Actor: The agent or service that proposed the decision, and the business domain it operates in.
Model: Exact model or model-routing target that produced the output, pinned by version.
Context hash: SHA-256 of the retrieved context bundle the model saw. Replay reloads context by this hash, not by query.
Guardrails: How many policy checks ran and the worst result: pass, flagged (escalated to a human) or blocked.
Outcome: What the system did: approved, executed, routed, escalated to review, or rejected.
Confidence: Calibrated score for the output. Below the 0.72 floor the decision escalates automatically.
Reviewer: The human who attested to the decision, when policy or a guardrail required one.
Ledger hash: SHA-256 over the previous record's hash and this record's canonical body — the link in the chain.
Parent: The upstream decision whose output this one consumed. Parent links form the lineage graph.

Put your decisions on a ledger

We instrument your existing agents and models with context hashing, guardrail verdict capture and a hash-chained decision ledger on WORM storage — then prove it with replay against your own records.