The field guide to controlled agentic commerce.
Understand what changes when software can buy, where authorization belongs, and how Mandate keeps humans and deterministic rules in control.
Build the right mental model first.
Start with the commerce journey and the questions operators ask before trusting an agent with purchasing work.
Field guideHow agentic commerce worksFollow one business purchase from human intent through agent action, authorization, payment infrastructure, and audit evidence.12 minPillar guideWhat are agentic payments?Understand the five layers behind an AI-agent payment: intent, identity, authorization, execution, and evidence.13 minFAQAgentic commerce and Mandate FAQsPlain answers about agents, mandates, approvals, API keys, risk, privacy, and the simulation boundary.24 answers
Understand the control layer.
Use these references when designing an integration, reviewing a decision, or deciding what an agent should be allowed to do.
ComparisonAI-agent payment platforms comparedCompare Mandate with Skyfire, Stripe Issuing, and Ramp by layer, controls, execution, approvals, and evidence.4 productsReferenceMandate knowledge baseA working reference for authorization decisions, spending mandates, scoped credentials, risk escalation, and audit records.6 topicsSecuritySecurity and trust boundarySee why the language model cannot authorize, how untrusted instructions are contained, and which production controls remain planned.Current model
Reason about the category.
Short arguments and explainers on the operating principles behind safe agent purchasing.
Protocol guideAP2 and the agent payments protocol, explainedA practical guide to typed mandates, signed intent, and protocol-level authorization evidence.11 minUse-case guideAI agents in procurement: where controls belongA business use case for budgets, suppliers, approval thresholds, and controlled purchasing authority.12 minPoint of viewAgentic commerce needs a separation of powersWhy the system interpreting a request should not be the system granting financial authority.7 minExplainerFrom user intent to an auditable purchase requestA concrete trace of what must be recorded before an agent-initiated purchase can be trusted.6 minDesign principleRisk scores should not override policyRisk is useful for escalation. It should never erase an explicit financial boundary.5 min