Inference Acceptance Layer · Demo

Bring any model. AiGentsy decides whether its output is allowed to become consequence.

LLM output should not automatically become action. The AiGentsy Acceptance Runtime routes model output through policy, evidence, and HoverStack reuse — and decides whether the downstream consequence is allowed, blocked, or held.

5 deterministic benchmark fixtures · raw_output hand-written (not a live LLM call) · HoverStack metrics labeled estimated · live cross-model benchmark (GPT / Claude / Gemini / Llama / Qwen) deferred to a controlled provider-benchmark harness pass.

Developers can post raw model output to /acceptance-runtime/evaluate to receive a runtime decision, evidence record, and export path. The runtime returns the same decision shape (accepted / rejected / retry / escalated) and consequence state (allowed / blocked / held) you see in the 3-lane comparison below.

Recall what was proven. Accept what is allowed. Prove what happened. Verify the record. Settle only when consequence is authorized.

Aggregate metrics across all 5 benchmark fixtures

Methodology defined in /data/inference_acceptance_scenarios.json · estimated metrics labeled est

Savings Trace · Before / After

What the acceptance gate prevented, reused, shortened, escalated, or verified.

The same five deterministic fixtures, viewed as a presentation layer over existing Acceptance Runtime, HoverStack reuse, ProofPack export, and verifier outputs. Nothing here is invented or persisted — every line below derives from a fixture or runtime field that already exists.

Deterministic demo fixtures only — no live LLM call. Live cross-model evaluation is the scope of the /acceptance-runtime/benchmarks operator harness. Every Savings Trace item is labeled measured, estimated, demo/reference, or provider-measured; default fixture labels are demo/reference + estimated.

Savings Trace shows what the gate prevented or reused. Consequence Memory shows what the chain records for proof, review, and future Recall. Same offline verifier as the handoff demo. Both are presentations of existing fields, not new artifacts. Open the Vault →

Watch a real settlement happen in 60 seconds

This demo replays real HoverStack decisions from CUDA-validated benchmark runs, driving live settlement through our production protocol. Settlement is one of several consequence types AiGentsy gates — deployment, handoff, API action, procurement, and inference acceptance follow the same accept-before-consequence pattern shown in the six held-consequence cards below and in the Inference Acceptance Layer above. The compute decisions shown are from actual Qwen2.5-7B inference runs. The ProofPack is cryptographically real and appears in our production Merkle log. Settlement fires through real test-mode Stripe. ProofPack Reuse, our benchmark-proven v1.7 mechanism, eliminates redundant compute when agents encounter already-attested work.

Creates a real demo agent, deal, and ProofPack on our production protocol.

State: INITIALIZED
HoverStack decisions (live replay from v1.6 benchmark)
0 / 54 decisions streamed
ProofPack Created
Deal ID
Proof Hash
Policy Hash
Scope Lock
The proof is verified. Accept to settle, or reject — consequence will be held and the signed rejection reason recorded.
Settlement Complete
Stripe PI
Amount$100.00 (test mode)
StateSETTLED
Acceptance rejected · Consequence held
VerifyPASS
AcceptREJECT
ConsequenceHELD
Dispute ID

Signed REJECTED event recorded in the Vault — reason and failed checks travel in the bundle. Dispute path opened.

View signed rejection record →
Download Proof Bundle Verify in Browser

The bundle is a real, offline-verifiable cryptographic artifact from our production Merkle log.

The demo above is the happy path: the agent works, a proof is created, the proof verifies, and the buyer accepts — so settlement fires. The scenario below is the edge case the wedge is built around. Cryptographic verification and acceptance are two different gates. A proof can be authentic, untampered, and traceable to the mandate, and the acceptance policy can still reject it — in which case settlement, release, deployment, or handoff is held. The signed REJECTED event with reason and failed checks is what makes the rejection auditable.

Scenario

Verified but Rejected

A proof can verify cryptographically and still fail acceptance. In this example, the proof bundle is valid, but the acceptance policy rejects because required checks are missing. Settlement or downstream action is held.

This demo mirrors the settlement-native-mcp starter policy fixture and adapter contract (acceptance_policy.example.json + adapter_contract.example.json). The three booleans below are runtime-compatible policy fields — the adapter output is validated by starter_boolean_validator into normalized_policy_inputs before acceptance evaluates. The signed bundle freezes the AdapterEvaluation (adapter_id, adapter_version, contract_hash, input_schema_hash, input_hash, output_hash, validation_result); bundle_hash binds it.

Mandate
Produce a release candidate with passing tests, rollback instructions, and reviewer approval before deployment.
Proof submitted
Signed proof bundle: artifact hash, timestamp, test summary.
Verify
PASS Proof bundle is authentic, untampered, traceable to the mandate.
Acceptance policy
tests_passed = true
rollback_plan_present = false
reviewer_approval = false
Accept
REJECT
Consequence
HELD No settlement, release, deployment, or handoff.
Signed rejection reason recorded
“Proof verified, but acceptance failed because rollback plan and reviewer approval were missing.”
View signed rejection record →

Each row in the Vault is a real signed REJECTED event with reason and failed checks — the bundle passes aigentsy-verify offline. Signed rejection records include the policy_snapshot and evaluated_inputs needed to replay the decision; the bundle hash binds every byte.

Want to register your own adapter contract? Start with aigentsy adapter scaffold --id your.adapter --version 0.1.0 --validator boolean, then lint it against the AdapterContract schema. Docs →

Held by acceptance

Six live test-mode consequence gates show the same invariant: proof can verify while acceptance fails and downstream consequence stays held. AdapterContracts validate signals into typed inputs. The counterparty defines the standard. AiGentsy enforces it.

"Verified but Rejected", "Payout held", "Deployment held", "Handoff held", "API Action held", and "Procurement held" are all backed by live starter policy fixtures — click Run on each card to drive a real bundle through the gate. Each scenario uses test-mode consequence semantics; no real money moves, no real deployment triggers, no real external handoff fires, no real API call is made, no real purchase order is created. Every exported bundle replays offline with aigentsy-verify.

The six live consequence gates
Verified but Rejected
Proof verifies. Acceptance rejects. Downstream action held.
Payout Held
payout_held · no_funds_moved=true
Deployment Held
deployment_held · no_deployment_triggered=true
Handoff Held
handoff_held · no_handoff_triggered=true
API Action Held
api_action_held · no_api_action_triggered=true
Procurement Held
procurement_held · no_purchase_order_created=true
Scenario — Payout held · runnable

Proof verifies, but PO mismatch + missing counterparty approval. Payout stays held.

Mandate
Pay contractor only if deliverable submitted, invoice present, PO matches, amount within mandate, and counterparty approval present.
Policy inputs
deliverable_submitted = true
invoice_present = true
po_match = false
amount_within_mandate = true
counterparty_approval = false
Accept / reject
REJECT — rule 0: po_match must equal true
Consequence
PAYOUT HELDdownstream_triggered=false · test-mode payout consequence · no funds moved
Replay
Exported bundle verifies offline with aigentsy-verify.
Scenario — Deployment held · runnable

Tests pass, but rollback plan missing + reviewer approval missing. Release stays held.

Mandate
Release candidate may deploy only if tests pass, artifact is present, rollback plan is present, deployment window is approved, and reviewer approval is present.
Policy inputs
tests_passed = true
artifact_present = true
rollback_plan_present = false
deployment_window_approved = true
reviewer_approval = false
Accept / reject
REJECT — rule 0: rollback_plan_present must equal true
Consequence
DEPLOYMENT HELDdownstream_triggered=false · test-mode deployment consequence · no deployment triggered
Replay
Exported bundle verifies offline with aigentsy-verify.
Scenario — Handoff held · runnable

Agent output exists, but checklist incomplete + counterparty approval missing. Handoff stays held.

Mandate
Work may be handed off only if the deliverable is submitted, the required checklist is complete, counterparty acceptance is present, and the downstream recipient is authorized.
Policy inputs
deliverable_submitted = true
checklist_complete = false
counterparty_approval = false
recipient_authorized = true
Accept / reject
REJECT — rule 0: checklist_complete must equal true
Consequence
HANDOFF HELDdownstream_triggered=false · test-mode handoff consequence · no handoff triggered
Replay
Exported bundle verifies offline with aigentsy-verify.
Scenario — API Action held · runnable

Agent requests an API call. Mandate doesn’t authorize the action and reviewer approval is missing. API action stays held.

Mandate
Agent may call a downstream API only if the action was requested, the target is allow-listed, the payload schema is valid, the mandate authorizes the action, and reviewer approval is present.
Policy inputs
api_action_requested = true
api_target_allowlisted = true
payload_schema_valid = true
mandate_authorizes_action = false
reviewer_approval = false
Accept / reject
REJECT — rule 0: mandate_authorizes_action must equal true
Consequence
API ACTION HELDdownstream_triggered=false · test-mode API action consequence · no external API call made
Replay
Exported bundle verifies offline with aigentsy-verify.
Scenario — Procurement held · runnable

Agent recommends a vendor. Budget exceeds mandate + procurement approval missing. Procurement stays held.

Mandate
Agent may recommend a vendor, but procurement may proceed only if the action was requested, the vendor is approved, budget is within mandate, procurement approval is present, and both compliance + conflict checks pass.
Policy inputs
procurement_requested = true
vendor_approved = true
amount_within_mandate = false
procurement_approval = false
compliance_check_passed = true
conflict_check_passed = true
Accept / reject
REJECT — rule 0: amount_within_mandate must equal true
Consequence
PROCUREMENT HELDdownstream_triggered=false · no purchase order created · no vendor commitment made · test-mode procurement consequence
Replay
Exported bundle verifies offline with aigentsy-verify.