AiGentsy — Evidence Vault

See the gate before consequence.

Logs observe. Gates decide. This is the interactive demo of AiGentsy Vault — the enterprise control-plane workspace powered by the AiGentsy Consequence Layer. It shows how AiGentsy gates consequential agent actions before they move downstream, then produces signed, verifiable evidence afterward. Run the native reference agent through mandate → work → evidence → acceptance decision → consequence state → ProofPack → Verify — then attach your own agent.

Mandate before work. Proof before acceptance. Acceptance before consequence. Settlement when value moves.

Unauthorized consequences are stopped before they happen; approved work proceeds with portable, independently verifiable evidence — while execution stays in your systems.

Recall· Accept· Prove· Verify· Settle

Demo / reference evidence only

What this Vault proves

L1 · Mandate

Mandate-bound work

Every record traces back to the adapter, policy, and contract that authorized it — upstream of the work itself.

Live

L3 · Proof

ProofPack evidence

Portable, signed, content-addressed bundles — hash chain, signed Merkle tree head, embedded key directory.

Live

L4 · Gate

Acceptance-gated consequence

A verified proof is not authorization. Payout / deployment / handoff / API action fires only after acceptance clears.

Live

L4 · Exact-auth

Exact authorization

Acceptance decides whether the consequence may proceed; exact authorization binds the approved action to configured consequence fields. Tier-1 Acceptance is the live guard; Tier-2 enforcement is set per deployment — shadow mode is not enforced Tier-2. Authorization metadata does not itself perform the external action.

Deployment-set

L6 · Execution

Enterprise-owned execution

The enterprise or its provider performs the external action. AiGentsy governs permission and holds the causal trail — never custody of credentials, funds, target systems, or result payloads.

Non-custodial

L7 · Verifier

Offline verification

Standalone aigentsy-verify==1.5.0 checks bundle hash, event chain, Merkle inclusion, signed STH — no runtime call.

Live

L8 · Reconcile

OutcomeReceipt vs reconciliation

OutcomeReceipt is a performer-signed outcome attestation. Reconciliation is a later caller-attributed observation — additive and non-overwriting — and does not by itself prove external-world truth.

Explanatory

One Vault. Every consequence path.

One Vault view for accepted, rejected, held, retried, escalated, settled, and verified autonomous work.

The Vault brings every AiGentsy branch into one enterprise surface: handoff, acceptance, refusal, inference gating, Consequence Memory, Savings Trace, ProofPack export, and verifier-backed evidence.

Run the verifier locally

Download the handoff bundle, save it as bundle.json, then run the standalone Python verifier. No AiGentsy account. No runtime call. No blockchain.

pip install aigentsy-verify==1.5.0 aigentsy-verify bundle bundle.json --fetch-key

The verifier checks bundle hash, event hash chain, Merkle inclusion, and the signed STH. spec_version=2.0.0 bundles validate at Tier-1 (per-event classification is attribution-only or platform-attested; never claimed as independently signed). spec_version=3.0.0 bundles additionally verify per-event actor_signatures against the embedded key_directory.

The signed record of autonomous work.

Mandate upstream. Proof at handoff. Acceptance before consequence. Settlement when value moves or work advances.

Every mandate, ProofPack, acceptance decision, refusal, outcome receipt, and downstream consequence — searchable, exportable, and independently verifiable. Each row links to its signed bundle and a copy-paste command your auditor can run offline.

Consequence workflow

Mandate› Agent work› Proposed consequence› Policy gate› Acceptance› Exact authorization› External execution› Outcome evidence› Reconciliation› Consequence Memory

AiGentsy governs the policy gate, the acceptance decision, and the causal trail. Only accepted work proceeds to authorized consequence. External execution is performed by the enterprise or its provider — AiGentsy does not perform the external action, and authorization metadata does not itself execute it.

Decision states · read-only

Accepted · allowed Rejected · blocked Escalated · held Retry · held

These are the decision states already recorded on the chain. This workspace displays them; it does not decide them. The four branches are demonstrated end-to-end in the scenario cockpit.

See these branches demonstrated → Open operator view

Non-custody boundary

AiGentsy holds the causal trail

Mandates and actor references
Policy inputs
Acceptance records
Authorization metadata
Hashes, identifiers, signatures, and event-chain evidence
ProofPacks
OutcomeReceipts
Reconciliation observations
Consequence Memory projections

Enterprise or provider owns and operates

Agents and models
Credentials and private keys
Compute and execution environments
Documents and customer artifacts
Target systems and callbacks
Funds and settlement assets
External result payloads
Provider accounts

Verification establishes integrity, provenance, signatures, and event-chain consistency — it does not independently prove external-world truth.

Refusals & Disputes

Completed Work

Consequence Records

See the gate before consequence.

What this Vault proves

One Vault. Every consequence path.

Six branches. One signed evidence surface.

Run the verifier locally

Map this to a consequential workflow.

The signed record of autonomous work.

AiGentsy holds the causal trail

Enterprise or provider owns and operates

Evidence ledger