What does an agentic commerce platform do?

It lets AI agents take part in transactions. Depending on type, it exposes merchant catalogs and checkout to agents, gives agents wallets or payment credentials with spending rules, or lets API providers charge agents per request. It also provides controls and records for users and merchants.

What protocols do agentic commerce platforms use?

Common ones include the Agentic Commerce Protocol for agent checkout with merchants, Google's Agent Payments Protocol for verifiable user mandates, and x402 for pay-per-request payments over HTTP, usually in stablecoins. Many platforms support more than one protocol to reach more counterparties.

What Is an Agentic Commerce Platform?

An agentic commerce platform is software that lets AI agents transact: it exposes products or services in machine-readable form, handles checkout or payment requests from agents, manages delegated payment credentials, and gives merchants and users controls and records. Some platforms serve merchants selling to agents; others give agents wallets and spending rules.

Merchant-side platforms

Merchant-side platforms help sellers reach agents. They publish product feeds or APIs agents can query, accept orders through agent checkout protocols, and process payment with tokens scoped to the transaction. Commerce platforms and payment processors have added features in this category.

Agent-side platforms

Agent-side platforms equip agents to pay. They issue wallets or payment credentials, enforce spending policies, sign or authorise transactions, and log activity. Some are custodial services holding funds for the agent; others are non-custodial, leaving keys with the developer or user.

API and data monetisation platforms

A third type lets API and data providers charge agents per call. These platforms meter usage and accept payment programmatically, sometimes with HTTP 402 based flows such as x402, so an agent can pay for a request without a prior account or subscription.

Core components

Most platforms include discovery or catalog interfaces, a payment layer, identity or authorisation checks, policy controls such as caps and allowlists, and reporting. The payment layer may use cards, bank transfers, or stablecoins depending on the use case.

Protocols supported

Check which protocols a platform supports. The Agentic Commerce Protocol targets agent checkout with merchants, the Agent Payments Protocol targets verifiable user mandates, and x402 targets pay-per-request over HTTP. Support for open protocols reduces lock-in.

Custody model

Custodial platforms hold funds and keys, which simplifies setup but adds counterparty risk and often compliance obligations. Non-custodial platforms keep keys with the user or developer and enforce limits cryptographically, for example with session keys that can only spend a bounded float.

Settlement

Card settlement takes days and carries chargeback risk; stablecoin settlement on a fast chain finalises in seconds and suits small amounts. The right choice depends on whether the counterparty is a consumer merchant or an API charging fractions of a dollar.

Controls and safety

Look for per-transaction and per-period spending caps, merchant or address allowlists, expiry on credentials, instant revocation, and complete logs. Agents can be manipulated, so controls must be enforced outside the model, not by instructions alone.

Building with Is an Agentic Commerce Platform

Good platforms offer SDKs, clear docs, test environments, and simple ways to attach payment capability to existing agent frameworks. Testnet support lets developers exercise payment flows without real funds.

How to choose

Match the platform to the job. Selling physical goods to consumer assistants points toward merchant-side checkout integrations. Paying for APIs and data from autonomous agents points toward programmatic stablecoin payments with bounded, non-custodial wallets.

Platform risks

Risks include vendor lock-in, unclear liability when agents misbehave, regulatory changes, and immature standards. Prefer platforms built on open protocols, with exportable records and clear terms about responsibility.

Where the category is heading

Expect convergence around a few open protocols, more merchants exposing agent-readable catalogs, and payment credentials designed specifically for delegated machine use. Platforms that interoperate across protocols will be easier to adopt.

Identity and agent verification

Merchants increasingly want to distinguish legitimate agents from bots and scrapers. Platforms address this with signed requests, registered agent identities, and verifiable mandates showing that a real user authorised a purchase. For machine payments, a valid onchain payment itself acts as proof that the caller is willing to pay for the service.

Reporting and reconciliation

Businesses need records that tie each agent payment to a task, a user, and an invoice or receipt. Good platforms export transaction logs with metadata, support accounting integrations, and make onchain transaction hashes easy to reference. Clear records simplify audits and disputes.

Pricing of platforms

Agentic commerce platforms charge in different ways: per transaction fees, percentage of volume, subscriptions, or free open source software with paid hosted services. For high-frequency, low-value machine payments, per-transaction fees can dominate costs, so model expected volume before choosing.

Questions to ask a vendor

Ask who holds keys and funds, how spending limits are enforced, which protocols and networks are supported, how fast settlement is, what happens if an agent is compromised, how revocation works, what records are available, and what fees apply at your expected volume.

Testing before going live

Before giving agents real funds, run payment flows on testnets or in sandbox modes with test credentials. Simulate failures such as rejected payments, price changes, and revoked keys. Testing how the platform behaves under these conditions reveals gaps in controls before they cost money.

Building versus buying

Teams can assemble their own agent payment stack from open protocols, wallet libraries, and payment providers, or adopt a platform that packages these pieces. Building offers control and avoids fees but takes engineering time and security expertise. Buying is faster but adds dependency on a vendor's roadmap and terms.

Frequently asked questions

What does an agentic commerce platform do?: It lets AI agents take part in transactions. Depending on type, it exposes merchant catalogs and checkout to agents, gives agents wallets or payment credentials with spending rules, or lets API providers charge agents per request. It also provides controls and records for users and merchants.
Are agentic commerce platforms custodial?: Some are and some are not. Custodial platforms hold funds and keys for agents. Non-custodial platforms keep keys with the user or developer and enforce limits cryptographically, for example with session keys that can only spend a bounded amount for a limited time.
What protocols do agentic commerce platforms use?: Common ones include the Agentic Commerce Protocol for agent checkout with merchants, Google's Agent Payments Protocol for verifiable user mandates, and x402 for pay-per-request payments over HTTP, usually in stablecoins. Many platforms support more than one protocol to reach more counterparties.