Online shopping has always assumed a human is clicking the buttons. Agentic commerce changes that assumption. In agentic commerce, an AI agent can search for products, compare options, fill a cart, and complete a purchase on a person’s behalf, within limits that person has set.
This guide explains what agentic commerce is, how it works, the main protocols behind it, and the risks businesses and shoppers should understand.
What Is Agentic Commerce?
Agentic commerce is the buying and selling of goods and services where an AI agent carries out some or all of the shopping process for a user. Instead of visiting several websites, a shopper might tell an assistant, “Find me running shoes under $120 in size 10 and buy the best-reviewed pair,” and the agent handles the rest.
The difference from earlier e-commerce tools is autonomy. A recommendation engine suggests products. A shopping agent can take action: it checks availability, applies the user’s preferences, and, with permission, pays.
If you are new to the idea of AI agents, start with our guide: AI Agents Explained.
How Agentic Commerce Works
Most agentic commerce flows follow the same basic steps:
- Intent: the user describes what they want, along with limits such as budget, brand, or delivery date.
- Discovery: the agent searches product catalogs and compares options.
- Cart and checkout: the agent builds an order and sends it to the merchant.
- Authorization: the user confirms the purchase, or a pre-approved rule allows the agent to proceed.
- Payment: a payment credential is used, often a token limited to a specific merchant and amount.
- Fulfillment and support: the merchant ships the order and handles returns, as with any normal sale.
The hardest part is not finding products. It is proving that the agent is acting with the user’s real permission, and making sure merchants, payment networks, and banks can trust the transaction.
Why Agentic Commerce Is Growing Now
Three developments pushed agentic commerce from an idea to real products in 2025 and 2026:
- Capable AI assistants: chat assistants now reach very large audiences. When OpenAI launched shopping checkout in ChatGPT in September 2025, it said more than 700 million people used ChatGPT every week.
- Open protocols: AI companies, payment providers, and retailers released shared standards so agents and stores can work together without one-off integrations.
- Payment network support: card networks introduced ways to issue credentials specifically for AI agents, with spending controls.
The Key Agentic Commerce Protocols
Several open standards now shape how agentic commerce works. They solve different parts of the problem, and many are designed to work together.
1. Agentic Commerce Protocol (ACP)
OpenAI and Stripe introduced the Agentic Commerce Protocol on September 29, 2025, alongside Instant Checkout in ChatGPT. At launch, U.S. users could buy from U.S. Etsy sellers directly in a chat, with Shopify merchants announced as coming soon.
Under ACP, the AI platform passes order details to the merchant’s own systems. The merchant decides whether to accept the order, processes payment through its existing provider, and remains the merchant of record. OpenAI said payment tokens are authorized only for specific amounts and specific merchants, and users confirm each step. According to the ACP website, the protocol is open source under the Apache 2.0 license and works with both REST and MCP.
2. Agent Payments Protocol (AP2)
Google announced the Agent Payments Protocol on September 16, 2025, saying it was developed with more than 60 organizations, including American Express, Mastercard, PayPal, Coinbase, Adyen, Etsy, and Salesforce. AP2 is payment-agnostic, covering cards, real-time bank transfers, and stablecoins.
AP2’s core idea is the “mandate”: a signed, verifiable record of what the user authorized. In Google’s original description, an intent mandate captures the user’s request, and a cart mandate records approval of the exact items and price. This creates an audit trail if something goes wrong.
On April 28, 2026, Google announced AP2 version 0.2, with support for “human not present” purchases based on pre-authorized instructions, and said it was donating AP2 to the FIDO Alliance so the standard remains community-led.
3. Universal Commerce Protocol (UCP)
The Universal Commerce Protocol describes itself as a common language for platforms, agents, and businesses. It covers the full shopping journey, including catalog, cart, checkout, identity linking, and orders. Its shopping specification lists co-developers including Google, Shopify, Etsy, Wayfair, Target, Walmart, Amazon, Microsoft, Meta, Salesforce, and Stripe.
UCP says it has built-in support for AP2 for payments, along with the A2A protocol and MCP. Lodging and food specifications are also in progress, with travel and delivery companies listed as co-developers.
4. Visa Intelligent Commerce
On April 30, 2025, Visa introduced Visa Intelligent Commerce, a set of APIs and a partner program for AI platforms. It replaces raw card details with tokenized credentials that confirm a shopper’s chosen agent is allowed to act, and lets consumers set spending limits. Visa named partners including Anthropic, IBM, Microsoft, Mistral AI, OpenAI, Perplexity, Samsung, and Stripe.
5. x402
Coinbase introduced x402 in May 2025. It uses the web’s long-unused “402 Payment Required” status code to let agents pay for online resources with stablecoins inside a normal web request. It is aimed at machine-to-machine payments, such as an agent paying per use for data or an API.
Agentic Commerce Protocols Compared
| Protocol | Led by | Main focus |
|---|---|---|
| ACP | OpenAI and Stripe | Checkout between AI platforms and merchants |
| AP2 | Google, now at the FIDO Alliance | Proving user authorization for agent payments |
| UCP | Industry group incl. Google, Shopify, Walmart | Full journey from discovery to order |
| Visa Intelligent Commerce | Visa | Tokenized card credentials for agents |
| x402 | Coinbase | Stablecoin payments inside web requests |
These standards are not all direct competitors. A single purchase could use UCP or ACP for checkout, AP2 to prove authorization, and a network token from a card scheme for the payment itself. Many build on the agent standards we covered in What Is the A2A Protocol? and MCP vs API.
Benefits of Agentic Commerce
For shoppers
- Less time comparing products across many sites.
- Purchases that follow personal rules, such as budget, brand, or sustainability preferences.
- Automated buying for routine items or time-sensitive purchases, where users allow it.
For merchants
- A new sales channel inside AI assistants.
- Structured orders that arrive ready to process.
- Under ACP and UCP, merchants keep their customer relationship and remain the merchant of record.
Risks and Challenges
Agentic commerce also creates new problems that the industry is still working through:
- Authorization and disputes: if an agent buys the wrong item, who is responsible? Signed mandates and audit logs are designed to answer this.
- Fraud: criminals may build fake agents or try to hijack real ones. Agent-specific tokens and verified identity are key defenses.
- Prompt injection: a product page or review could hide instructions that try to steer an agent. Read our guide to prompt injection to understand the risk.
- Access disputes: some institutions are cautious about letting outside agents act on customer accounts, as covered in our report on banks blocking third-party AI agents.
- Fragmented standards: several protocols exist at once, and merchants may need to support more than one.
How Businesses Can Prepare for Agentic Commerce
- Clean up product data: agents rely on accurate titles, prices, stock levels, and attributes. Structured product feeds matter more than ever.
- Check your platform and payment provider: commerce platforms and payment processors are adding agentic checkout support. Ask what they offer.
- Review policies: make return, refund, and dispute rules clear for orders placed by agents.
- Strengthen fraud controls: plan how you will recognize legitimate agents and reject suspicious ones.
- Start small: test with a limited product range before expanding.
Conclusion
Agentic commerce moves AI from recommending products to actually buying them, with the user’s permission. The building blocks are already in place: ACP and UCP for checkout, AP2 and network tokens for trusted payment authorization, and x402 for machine payments. The next stage will depend on trust, including clear consent, strong fraud protection, and fair rules when something goes wrong.
Frequently Asked Questions
What is agentic commerce in simple terms?
Agentic commerce is shopping where an AI agent finds, selects, and buys products for you, following the rules and limits you set.
Is agentic commerce safe?
The main protocols include safeguards such as user confirmation, payment tokens limited to specific merchants and amounts, and signed records of what the user approved. Risks such as fraud and prompt injection still exist, so caution is important.
What is the difference between ACP and AP2?
ACP, from OpenAI and Stripe, focuses on the checkout between an AI platform and a merchant. AP2, from Google, focuses on proving that the user authorized the payment, using signed mandates.
Do merchants lose the customer relationship in agentic commerce?
Under ACP and UCP, the merchant remains the merchant of record and handles payment, fulfillment, and support.
Can an AI agent buy something without asking me?
Only if you allow it. Standards such as AP2 version 0.2 support pre-authorized “human not present” purchases, but these are based on limits the user sets in advance.
The Future AI Wire Editorial Team covers artificial intelligence, AI agents, agentic AI, and emerging technologies. We publish clear, research-informed explanations and practical insights to help readers understand AI developments, real-world applications, and the technologies shaping the future.