Building the trust layer for the next generation of commerce
AI agents will initiate trillions in transactions. We build the real-time decision verification infrastructure between autonomous agents and the financial system.
One API call. Seven decision gates.
Send an authorization request and watch the Decision Trust Protocol evaluate it in real time — every gate, every score, every millisecond.
import mandate response = mandate.authorizations.create( agent_id="agent_procurement_01", amount="2400.00", currency="USD", merchant="mch_aws_7291", intent_context={ "intent_type": "PURCHASE", "task_reference": "TASK-2847", "reasoning_summary": "Scaling GPU instances for Q3 training run. Compared 3 providers on cost, latency. AWS optimal at $2,400.", "confidence": 0.94, "alternatives_considered": 3 } )
Nine modules. One trust layer.
Everything an issuer needs to authorize, monitor, and govern AI agents operating on existing payment rails — from real-time authorization to compliance reporting.
Agent Authorization
Seven-gate verification pipeline processes every transaction in 23ms. Intent evaluation, budget enforcement, sanctions screening, and behavioral checks.
Trust Scoring (KYA)
Agent Behavioral Fingerprint — 42 features, six per signal class across seven classes — with progressive trust levels. Drift detection catches compromised agents before fraud occurs.
Transaction Monitoring
Live feed of every authorization decision with KYA scores, risk signals, and one-click investigation. Full pipeline trace on every decision.
Compliance Reporting
SAR/STR narrative generation in FinCEN BSA and Panama UAF Law 23 format. Compliance calendar with regulatory deadline tracking.
Agent Registry
Complete lifecycle management for every AI agent. Trust levels, behavioral baselines, drift alerts, and multi-tenant isolation.
Ivy, the Compliance Agent
Natural language investigation and insights grounded in your transaction data. Draft SAR narratives, investigate declines, get proactive alerts.
Fraud Analytics
Agent-specific fraud detection: prompt injection, credential replay, behavioral drift attacks, and cross-agent correlation.
Risk Rules Engine
Visual rule builder with dual-approval governance, backtesting, and gap analysis that finds coverage holes before attackers do.
Case Management
Structured investigation workflows from alert to resolution. Auto-created cases, SLA tracking, and one-click SAR filing.
The tools your compliance, risk, and fraud teams need
Agent commerce introduces new operational challenges for compliance, risk, and fraud departments. Every authorization decision made by the protocol is fully explainable — your team can trace exactly why an agent was approved, declined, or escalated, what behavioral signals were active, and what the behavioral trust state was at the moment of decision. No black boxes. Every score has a reason. Every reason has an audit trail.
Transaction monitoring
Live feed of every authorization decision with KYA scores, risk signals, and one-click investigation. Filter by agent, merchant, amount, status, or time range. Drill into any transaction to see the full pipeline trace — which gates passed, which flagged, and the reasoning behind each score. Flag suspicious patterns and escalate directly from the monitoring view.
Agent registry & behavioral profiles
Complete visibility into every AI agent operating under your program. Track trust levels (REGISTERED through TRUSTED), behavioral risk zone (GREEN to CRITICAL), transaction history, decline rates, and behavioral baselines. Promote, throttle, or suspend agents with full audit trail. Understand how each agent's decision quality evolves over time.
Compliance reporting & regulatory export
Generate SAR narratives and STR reports in FinCEN BSA format (US) and Panama UAF Law 23 format (LatAm). Automated compliance calendar with regulatory deadline tracking. Export authorization logs in CSV or JSON for external auditors. Every report includes the complete decision context — intent type, reasoning summary, risk signals, and KYA score at the time of transaction.
Fraud analytics & case management
Identify anomalous patterns across agents, merchants, and time periods. The dashboard surfaces eight signal categories drawn from the protocol’s seventeen-type risk taxonomy — velocity spikes, geographic anomalies, amount outliers, behavioral drift, and more. Flag cases directly from the dashboard, assign to investigators, and track disposition through resolution. Chargeback management with evidence collection and Mastercard dispute lifecycle tracking.
Seven decision gates. Two system steps. Twenty-three milliseconds.
Every authorization traverses the full Decision Trust Protocol pipeline — from agent resolution to post-decision analysis.
Agent resolve
Identify the agent, load its mandate, trust tier, and behavioral profile from the registry.
Intent verification
Parse the intent context and validate reasoning quality, confidence calibration, and task alignment.
Anomaly gate
Run six-dimensional anomaly detection against the agent's behavioral baseline using EDQS scoring.
Mandate enforcement
Verify the transaction falls within the agent's spending limits, merchant allowlists, and time-of-day policies.
Behavioral overlay
Assess the agent's current behavioral state — detect reasoning degradation, drift, and prompt injection before they reach the network.
Risk scoring
Compute composite risk across merchant reputation, transaction patterns, velocity, and geographic signals.
KYA decision
Generate the final Know Your Agent trust score and map to GREEN, AMBER, RED, or CRITICAL trust zone.
Persist + webhooks (system)
Write the full decision record to the audit log. Fire webhooks for downstream systems and compliance tools.
Async post-decision (system)
Enqueue behavioral model updates, regulatory checks, and cross-agent correlation analysis.
Advancing the science of agent trust
We believe AI agents will become the primary participants in global commerce. Our research mission is to accelerate this transition on the safest possible path — building the verification frameworks, behavioral models, and trust protocols that make autonomous financial participation reliable at scale. Every model we train, every anomaly detector we design, and every standard we propose exists to solve one question: how do you trust a machine to spend money?
Economic Decision Quality Score
A six-dimension composite metric for evaluating the quality of AI agent economic reasoning, grounded in process-based supervision, Constitutional AI, and behavioral economics. Introduces a dual-audience architecture with pre-registered research hypotheses.
Session fingerprinting & prompt injection detection
Continuous behavioral monitoring that builds behavioral profiles for each agent within a single task session. Detects prompt injection attempts, reasoning degradation, and context manipulation in real time — catching behavioral shifts at the first anomalous transaction, before mandate enforcement even runs.
Where Mandate Labs sits in the flow
Two integration modes, one protocol. Both return the same enriched decision — the difference is when DTP is consulted.
Inside the authorization window
Your processor calls DTP during authorization decisioning. The enriched, advisory response — KYA score, risk signals, reason codes — returns inside the auth window and feeds your decision engine.
Before the transaction exists
Your agent platform calls DTP pre-presentment — the agent declares intent, DTP evaluates, and only approved actions become card transactions.
Sandbox access is self-serve. Production issuer deployments are sales-led and gated by Mandate Certified Integration — book a demo.
Built on payment industry foundations
The networks are building the identity and consent layer — Visa Intelligent Commerce and Trusted Agent Protocol, Mastercard Agent Pay’s Agentic Tokens, Google AP2 mandates, agentic checkout protocols. DTP consumes those identity signals and scores what no token scheme provides: the quality of the decision behind the transaction. Identity and consent travel with the token; decision trust travels with DTP.
Card-rail native messages
Card rails run ISO 8583 — decision payloads map to ISO 8583 authorization data elements for inline issuer integration, with ISO 20022 alignment for reporting and account-to-account flows.
Issuer RBA enrichment
KYA scores can enrich an issuer’s own risk-based authentication decisioning (ACS) for agent-initiated transactions — consumed issuer-side, alongside EMV 3-D Secure data.
DE 48.75 compatibility
Decision payloads — risk score, up to four reason codes, condition codes — are structured to map onto Mastercard DE 48.75 fraud-scoring conventions, and designed to coexist with network agent-identity standards as they ratify.
Production infrastructure, not a proof of concept
Built for the reliability, compliance, and support requirements of financial institutions.
Availability SLA
Multi-region deployment with automatic failover. Target 99.99% for enterprise tier.
Isolation architecture
Fully isolated tenant environments with dedicated encryption keys and data residency controls.
Compliance export
Automated suspicious activity reporting in FinCEN BSA and Panama UAF Law 23 formats.
Support & success
Slack Connect channel, named account manager, and quarterly business reviews.
Financial guarantees
Contractual uptime commitments with automatic service credits for latency or availability misses.
API stability
Semantic versioning with 12-month deprecation windows. No breaking changes without migration support.
The infrastructure for agent commerce is being built now
Join the institutions shaping the trust layer for autonomous transactions.
Get started