Get started

Plugin · governance and evidence

Decision Evidence

Structured, tamper-evident records of every AI decision. Capture what information agents considered, what tools they invoked, what decision they reached, and why. Cryptographically linked to the VOLT ledger and built on the AEE protocol. Enterprise compliance without retrofitting.

Installable plugin Enterprise tier Licence key unlock
$599per year · enterprise add-onAdd to cartView in store
LIVEdecision#00007chainVERIFIEDlevelfullhover a stage · the record is live
DECISION
Agent decides
An agent weighs its inputs and reaches a decision.
CAPTURED
Evidence record
Inputs, tools, reasoning and confidence, embedded in the AEE envelope.
CHAINED
VOLT ledger
Written to the ledger, hash-linked to the entry before it.
WITNESSED
WARD receipt
A witness receipt proves the record existed at that moment.
DEFENSIBLE #00007
sha256:9f2a…c4e1
SEALED ✓

CAPTURED · the evidence record

inputsalert.severity: critical · host: nw-worker-01
toolsfleet_status · service_health
decisionrestart_service · confidence 0.92
verifyChain(genesis → tip #00007) → TRUE
0 breaks · every entry hash-linked · WARD witness receipt attached

Watch one decision become sealed, defensible evidence

In plain words

What it is, where it lives, when to reach for it

What is it
A plugin for the QuoxCORE dashboard that records what each agent decision used: inputs, tools, reasoning and confidence.
Where do I use it
In the dashboard, attached to each agent reply's activity record, and on the VOLT and WARD pages.
When would I use it
When an auditor or a customer asks why an agent did something and you need the record.
How do I use it
Add it from the store; an admin enables it and every decision is recorded from then on, no agent code changes.

QuoxCORE is the free, self-hosted platform underneath this. What is QuoxCORE

How it installs: subscribe, paste the licence key into your QuoxCORE install, and it switches on. Nothing new to deploy, and no changes to your agents' code.

What it does

Every decision, recorded.

When an agent classifies a ticket, triggers a workflow, or takes an operational action, the plugin captures a structured record: what inputs the agent considered, what context it referenced, what tools it used, what it decided, and why.

  • Structured evidence in every AEE envelope. Evidence is embedded directly in the envelope, not bolted on after the fact.
  • Four levels: minimal, standard, full and none. Balance detail against token cost, with per-agent overrides for high-risk agents.
  • Automatic capture with no agent code changes. Switch it on and every decision is recorded from that point forward.
  • Confidence scoring on every decision. Set thresholds to flag low-confidence decisions for human review.
  • Tamper-evident by default. Every record is written to the VOLT ledger and hash-linked to the entry before it. Modifying any record breaks the chain.
decision evidence record
{
  "inputs_used": [
    "alert.severity: critical",
    "host: nw-worker-01",
    "service: nginx"
  ],
  "tools_used": [
    "fleet_status",
    "service_health"
  ],
  "decision": "restart_service",
  "reason_summary": "nginx on nw-worker-01 is unresponsive. No maintenance window active. Runbook authorises restart.",
  "action_taken": "systemctl restart nginx",
  "confidence": 0.92
}

Evidence capture

Structured records of inputs, reasoning, actions and confidence for every agent decision. No code changes required.

VOLT ledger

Automatic hash-linked recording to the Verifiable Operations Ledger. Tamper-evident by construction.

Decision timeline

Chronological view of all agent decisions with filtering, drill-down and chain traversal.

Evidence search

Query decision records by agent, time range, confidence, action type or free text across the full evidence corpus.

Audit export

Export evidence chains in structured formats for external auditors, compliance reviews or incident reports.

Token analytics

Track evidence token cost per agent, per level. Optimise the balance between detail and efficiency.

WARD witnesses

External attestation via Write-once Append-only Receipt Digests. Prove evidence existed at a point in time.

Confidence scoring

Every decision includes a confidence assessment. Set thresholds to flag low-confidence decisions for human review.

Per-agent levels

Apply full evidence to high-risk agents, minimal to routine ones. Granular control without global overhead.

Enterprise compliance

EU AI Act Article 12, NIST AI RMF, Colorado SB 21-169. Evidence infrastructure that satisfies regulatory requirements.

AEE integration

Evidence lives inside AEE envelopes as a first-class field. Not bolted on. Part of the protocol.

Incident reconstruction

Trace any operational outcome back through the decision chain. See what the agent knew, what it chose, and why.

How it installs

One key, no retrofitting.

Decision Evidence is an installable plugin. It unlocks with a licence key inside the QuoxCORE install you already run. There is nothing new to deploy.

01

An annual licence

Pick it up in the store for $599 a year. Checkout opens on launch day, through Stripe, and the licence key is delivered by email with the order.
02

Paste the key

Open your QuoxCORE dashboard, go to plugins, and paste the key. The plugin switches on without a rebuild or a restart.

03

Capture begins

Every agent decision is recorded from that moment. Choose an evidence level per agent: minimal, standard, full or none.

Pricing

An annual licence, renewed each year.

Installable plugin Enterprise tier

Decision Evidence

$599per year · enterprise add-on

One annual charge for the full plugin, with no per-decision charge on top. It runs on your infrastructure and records as many decisions as your agents make. Checkout opens on launch day, through Stripe, with your licence key delivered by email.

  • Evidence capture on every agent decision, no code changes
  • VOLT ledger recording with WARD witness receipts
  • Decision timeline, evidence search and audit export
  • Four evidence levels with per-agent overrides
  • Confidence scoring and incident reconstruction

Governed by design

Defensible to an auditor.

This plugin is the evidence story. Every material decision your agents make lands on the WARD chain: append-only, hash-linked, witnessed. When an auditor or a regulator asks why an agent did what it did, you do not tell a story. You hand over the chain and they check it themselves.

It is the same principle behind Git commit hashes and certificate transparency logs, applied to AI operational decisions. No blockchain overhead, no external dependencies, stored in your own infrastructure.

For auditors
verifyChain(genesis → tip) → TRUE
entry #7 · agent.decision.recorded
prev sha256:3b7d…a8f2 · 0 breaks
WARD receipt wr_2026030801 · confirmed

alter one byte → verifyChain() = FALSE