Table of Contents
What Is Agentic Commerce? The Definitive Guide to AI-Powered Purchasing
Arjun Bhargava
Co-founder and CEO @ Rye
Feb 20, 2026
15 minutes read
Agentic commerce is when AI agents don't just recommend — they buy. Learn how it works, where the market stands in 2026, and why checkout is the hardest problem.
TL;DR / Key Takeaways
Agentic commerce is what happens when AI agents don't just recommend products — they research, compare, and buy them autonomously on your behalf.
The market is projected to grow from $135 billion in 2025 to $1.7 trillion by 2030, with analysts estimating 15–25% of e-commerce transactions will be agent-driven by the end of the decade.
The biggest unsolved problem isn't discovery or recommendations — it's checkout. Merchant fraud detection systems block automated transactions, which is why most "AI shopping" demos quietly rely on humans behind the scenes.
New protocols from Google (Universal Commerce Protocol) and OpenAI (Agentic Commerce Protocol) are standardizing how agents interact with merchants, but both require merchant opt-in — leaving billions of products inaccessible to agents.
Universal checkout infrastructure, like Rye's Universal Checkout API, solves this by enabling agents to purchase from any merchant using just a product URL — no integration required.
Agentic commerce is what happens when AI agents don't just recommend products — they research, compare, and buy them autonomously on your behalf.
The market is projected to grow from $135 billion in 2025 to $1.7 trillion by 2030, with analysts estimating 15–25% of e-commerce transactions will be agent-driven by the end of the decade.
The biggest unsolved problem isn't discovery or recommendations — it's checkout. Merchant fraud detection systems block automated transactions, which is why most "AI shopping" demos quietly rely on humans behind the scenes.
New protocols from Google (Universal Commerce Protocol) and OpenAI (Agentic Commerce Protocol) are standardizing how agents interact with merchants, but both require merchant opt-in — leaving billions of products inaccessible to agents.
Universal checkout infrastructure, like Rye's Universal Checkout API, solves this by enabling agents to purchase from any merchant using just a product URL — no integration required.
Agentic commerce is what happens when AI agents don't just recommend products — they research, compare, and buy them autonomously on your behalf.
The market is projected to grow from $135 billion in 2025 to $1.7 trillion by 2030, with analysts estimating 15–25% of e-commerce transactions will be agent-driven by the end of the decade.
The biggest unsolved problem isn't discovery or recommendations — it's checkout. Merchant fraud detection systems block automated transactions, which is why most "AI shopping" demos quietly rely on humans behind the scenes.
New protocols from Google (Universal Commerce Protocol) and OpenAI (Agentic Commerce Protocol) are standardizing how agents interact with merchants, but both require merchant opt-in — leaving billions of products inaccessible to agents.
Universal checkout infrastructure, like Rye's Universal Checkout API, solves this by enabling agents to purchase from any merchant using just a product URL — no integration required.
Agentic Commerce, Defined
For two decades, e-commerce innovation mostly meant better search, better recommendations, and faster shipping. The purchase itself — the act of entering your card number, picking a shipping option, and clicking "buy" — stayed stubbornly manual.
That's changing. Fast.
Agentic commerce represents a fundamental shift in how goods and services are bought and sold online. Instead of a human navigating a website, comparing prices across tabs, and manually completing checkout, an AI agent handles the entire workflow: discovering products, evaluating options, calculating true costs, and executing the purchase — autonomously, in seconds.
This isn't a chatbot suggesting you might like a blue sweater. This is an AI system that knows your preferences, monitors prices, and buys the sweater when it hits your target price — while you're asleep.
The distinction matters. Previous generations of retail AI were reactive. Recommendation engines waited for you to browse. Chatbots waited for you to ask. Agentic AI is proactive. It plans, executes, and adapts — handling multi-step purchasing workflows that used to require a human at every turn.
Why Now? The Infrastructure Moment
Agentic commerce isn't new as a concept. Researchers have been writing about autonomous purchasing agents since the late 1990s. What's new is that the enabling infrastructure is finally catching up to the vision.
Three things converged in 2025 that made agentic commerce viable at scale:
Large language models got good enough to navigate the open web
Earlier automation tools relied on brittle scripts that broke whenever a website changed its layout. Modern LLM-powered agents can interpret page structure, adapt to unfamiliar checkout flows, and recover from errors — the same way a human would, but faster.
Protocols emerged to standardize agent-merchant interactions
The Model Context Protocol (MCP), Google’s Agent-to-Agent (A2A) framework, and commerce-specific standards like OpenAI’s Agentic Commerce Protocol and Google’s Universal Commerce Protocol created a common language for agents, merchants, and payment providers to communicate. Think of them as the HTTP of AI shopping — the plumbing that lets different systems talk to each other.
Payment infrastructure adapted for machine-to-machine transactions
Companies like Visa, PayPal, Stripe, and Lithic started building toolkits specifically designed for agent-initiated transactions — tokenized payment methods, programmable cards, and fraud-mitigation layers that distinguish legitimate agent activity from actual fraud.
The result: for the first time, an AI agent can go from "find me running shoes under $120" to a confirmed, shipped order without a human touching anything.
Agentic commerce isn't new as a concept. Researchers have been writing about autonomous purchasing agents since the late 1990s. What's new is that the enabling infrastructure is finally catching up to the vision.
Three things converged in 2025 that made agentic commerce viable at scale:
Large language models got good enough to navigate the open web
Earlier automation tools relied on brittle scripts that broke whenever a website changed its layout. Modern LLM-powered agents can interpret page structure, adapt to unfamiliar checkout flows, and recover from errors — the same way a human would, but faster.
Protocols emerged to standardize agent-merchant interactions
The Model Context Protocol (MCP), Google’s Agent-to-Agent (A2A) framework, and commerce-specific standards like OpenAI’s Agentic Commerce Protocol and Google’s Universal Commerce Protocol created a common language for agents, merchants, and payment providers to communicate. Think of them as the HTTP of AI shopping — the plumbing that lets different systems talk to each other.
Payment infrastructure adapted for machine-to-machine transactions
Companies like Visa, PayPal, Stripe, and Lithic started building toolkits specifically designed for agent-initiated transactions — tokenized payment methods, programmable cards, and fraud-mitigation layers that distinguish legitimate agent activity from actual fraud.
The result: for the first time, an AI agent can go from "find me running shoes under $120" to a confirmed, shipped order without a human touching anything.
Agentic commerce isn't new as a concept. Researchers have been writing about autonomous purchasing agents since the late 1990s. What's new is that the enabling infrastructure is finally catching up to the vision.
Three things converged in 2025 that made agentic commerce viable at scale:
Large language models got good enough to navigate the open web
Earlier automation tools relied on brittle scripts that broke whenever a website changed its layout. Modern LLM-powered agents can interpret page structure, adapt to unfamiliar checkout flows, and recover from errors — the same way a human would, but faster.
Protocols emerged to standardize agent-merchant interactions
The Model Context Protocol (MCP), Google’s Agent-to-Agent (A2A) framework, and commerce-specific standards like OpenAI’s Agentic Commerce Protocol and Google’s Universal Commerce Protocol created a common language for agents, merchants, and payment providers to communicate. Think of them as the HTTP of AI shopping — the plumbing that lets different systems talk to each other.
Payment infrastructure adapted for machine-to-machine transactions
Companies like Visa, PayPal, Stripe, and Lithic started building toolkits specifically designed for agent-initiated transactions — tokenized payment methods, programmable cards, and fraud-mitigation layers that distinguish legitimate agent activity from actual fraud.
The result: for the first time, an AI agent can go from "find me running shoes under $120" to a confirmed, shipped order without a human touching anything.
How Agentic Commerce Actually Works
At a high level, an agentic commerce transaction follows a loop that looks something like this:
1. Intent capture. The agent receives a goal — either explicitly ("order more printer paper") or through learned behavior (noticing you're running low based on past purchase cadence).
2. Research and discovery. The agent searches across merchants, compares prices, checks availability, reads reviews, and evaluates shipping options. Unlike a human who might check three or four sites, an agent can evaluate dozens in seconds.
3. Offer resolution. The agent needs to determine the true landed cost — product price plus tax plus shipping — for a specific buyer, at a specific address, from a specific merchant. This requires interacting with each merchant's checkout system in real time, because prices, taxes, and shipping rates aren't static. They vary by location, cart contents, and time of day.
4. Decision and purchase. Based on the resolved offers, the agent selects the best option (according to whatever criteria the buyer has defined — cheapest, fastest, most sustainable) and completes the transaction using a tokenized payment method.
5. Post-purchase management. The agent tracks the order, handles any issues, processes returns if needed, and feeds the outcome back into its model for future decisions.
Each step sounds simple. In practice, steps 3 and 4 are where almost every implementation breaks down.
At a high level, an agentic commerce transaction follows a loop that looks something like this:
1. Intent capture. The agent receives a goal — either explicitly ("order more printer paper") or through learned behavior (noticing you're running low based on past purchase cadence).
2. Research and discovery. The agent searches across merchants, compares prices, checks availability, reads reviews, and evaluates shipping options. Unlike a human who might check three or four sites, an agent can evaluate dozens in seconds.
3. Offer resolution. The agent needs to determine the true landed cost — product price plus tax plus shipping — for a specific buyer, at a specific address, from a specific merchant. This requires interacting with each merchant's checkout system in real time, because prices, taxes, and shipping rates aren't static. They vary by location, cart contents, and time of day.
4. Decision and purchase. Based on the resolved offers, the agent selects the best option (according to whatever criteria the buyer has defined — cheapest, fastest, most sustainable) and completes the transaction using a tokenized payment method.
5. Post-purchase management. The agent tracks the order, handles any issues, processes returns if needed, and feeds the outcome back into its model for future decisions.
Each step sounds simple. In practice, steps 3 and 4 are where almost every implementation breaks down.
At a high level, an agentic commerce transaction follows a loop that looks something like this:
1. Intent capture. The agent receives a goal — either explicitly ("order more printer paper") or through learned behavior (noticing you're running low based on past purchase cadence).
2. Research and discovery. The agent searches across merchants, compares prices, checks availability, reads reviews, and evaluates shipping options. Unlike a human who might check three or four sites, an agent can evaluate dozens in seconds.
3. Offer resolution. The agent needs to determine the true landed cost — product price plus tax plus shipping — for a specific buyer, at a specific address, from a specific merchant. This requires interacting with each merchant's checkout system in real time, because prices, taxes, and shipping rates aren't static. They vary by location, cart contents, and time of day.
4. Decision and purchase. Based on the resolved offers, the agent selects the best option (according to whatever criteria the buyer has defined — cheapest, fastest, most sustainable) and completes the transaction using a tokenized payment method.
5. Post-purchase management. The agent tracks the order, handles any issues, processes returns if needed, and feeds the outcome back into its model for future decisions.
Each step sounds simple. In practice, steps 3 and 4 are where almost every implementation breaks down.
Agentic Commerce vs. Traditional E-Commerce: What's Different?
It's worth being precise about what separates agentic commerce from the e-commerce automation that already exists. Coupon-clipping browser extensions, price-tracking tools, and "buy again" buttons on Amazon are all forms of shopping automation. But they're not agentic.
The difference comes down to autonomy and scope. A price tracker watches one product on one site and alerts you when it drops. An agentic system watches hundreds of products across dozens of sites, evaluates which one best fits your criteria right now, and completes the purchase — all without you intervening.
Traditional e-commerce is human-directed at every step: you search, you browse, you compare, you buy. Conversational commerce (chatbots, voice assistants) reduced friction at the discovery stage but still required human confirmation at checkout. Agentic commerce is the first model where the AI can close the loop end-to-end — from intent to delivered order — within defined guardrails.
That guardrail piece is critical. Nobody wants a rogue agent spending $5,000 on organic dog treats. Agentic commerce systems are designed with spending limits, approval thresholds, preference constraints, and audit trails that keep the human in control of policy while the agent handles execution.
It's worth being precise about what separates agentic commerce from the e-commerce automation that already exists. Coupon-clipping browser extensions, price-tracking tools, and "buy again" buttons on Amazon are all forms of shopping automation. But they're not agentic.
The difference comes down to autonomy and scope. A price tracker watches one product on one site and alerts you when it drops. An agentic system watches hundreds of products across dozens of sites, evaluates which one best fits your criteria right now, and completes the purchase — all without you intervening.
Traditional e-commerce is human-directed at every step: you search, you browse, you compare, you buy. Conversational commerce (chatbots, voice assistants) reduced friction at the discovery stage but still required human confirmation at checkout. Agentic commerce is the first model where the AI can close the loop end-to-end — from intent to delivered order — within defined guardrails.
That guardrail piece is critical. Nobody wants a rogue agent spending $5,000 on organic dog treats. Agentic commerce systems are designed with spending limits, approval thresholds, preference constraints, and audit trails that keep the human in control of policy while the agent handles execution.
It's worth being precise about what separates agentic commerce from the e-commerce automation that already exists. Coupon-clipping browser extensions, price-tracking tools, and "buy again" buttons on Amazon are all forms of shopping automation. But they're not agentic.
The difference comes down to autonomy and scope. A price tracker watches one product on one site and alerts you when it drops. An agentic system watches hundreds of products across dozens of sites, evaluates which one best fits your criteria right now, and completes the purchase — all without you intervening.
Traditional e-commerce is human-directed at every step: you search, you browse, you compare, you buy. Conversational commerce (chatbots, voice assistants) reduced friction at the discovery stage but still required human confirmation at checkout. Agentic commerce is the first model where the AI can close the loop end-to-end — from intent to delivered order — within defined guardrails.
That guardrail piece is critical. Nobody wants a rogue agent spending $5,000 on organic dog treats. Agentic commerce systems are designed with spending limits, approval thresholds, preference constraints, and audit trails that keep the human in control of policy while the agent handles execution.
Where Agentic Commerce Checkout Stands Today
Every major AI platform is racing to close the loop on shopping — moving from product recommendations to completed purchases inside the chat interface. Here's where the leading players stand as of early 2026.
OpenAI launched Instant Checkout in ChatGPT in February 2026, powered by the Agentic Commerce Protocol (ACP) — an open standard co-developed with Stripe. Users can now discover and purchase products without leaving the conversation. Product results are organic and unsponsored. Etsy is live, over 1 million Shopify merchants are in the onboarding pipeline, and PayPal has adopted ACP to expand its merchant coverage. With an estimated 50 million shopping-related queries per day, ChatGPT is quickly becoming the largest AI-native commerce channel.
Google introduced agentic checkout in AI Mode in late 2025, enabling purchases directly within search results for eligible merchants. The feature works with merchants who support guest checkout and Google Pay — including retailers like Wayfair, Chewy, and select Shopify stores. At NRF 2026, Google also unveiled the Universal Commerce Protocol (UCP), co-developed with Shopify and backed by over 20 partners including Visa, Mastercard, Target, and Walmart.
Perplexity expanded its Buy with Pro shopping feature to all users in November 2025 through a PayPal partnership, connecting shoppers to 5,000+ merchants with in-chat checkout.
Amazon launched its Buy for Me feature, allowing users to purchase from third-party retailers without leaving the Amazon app. The scope has expanded from roughly 65,000 to over 500,000 products, though it remains limited to single-item orders and doesn’t carry Amazon’s A-to-Z guarantee.
These are real, working products — not demos. The progress over the past twelve months has been remarkable.
The Universality Gap
But there's a catch. Every approach listed above depends on merchant participation. OpenAI's ACP requires merchants to implement the protocol and provide a product feed. Google's agentic checkout requires guest checkout and Google Pay support. Perplexity's in-chat purchases work with its partner merchants. Amazon's Buy for Me covers a curated catalog.
That leaves millions of online merchants — independent stores on custom platforms, niche retailers, international sellers — out of reach for any agent.
This is the core infrastructure challenge of agentic commerce: the open web has millions of merchants, but the checkout systems that AI agents can programmatically complete cover only a fraction of them. Merchant fraud detection systems (bot protection, CAPTCHAs, device fingerprinting, velocity checks) were built to block exactly the kind of automated transactions that agents need to perform. Until checkout infrastructure can work across any merchant — not just those who've opted into a specific protocol — the promise of truly universal agentic commerce remains partially fulfilled.
Every major AI platform is racing to close the loop on shopping — moving from product recommendations to completed purchases inside the chat interface. Here's where the leading players stand as of early 2026.
OpenAI launched Instant Checkout in ChatGPT in February 2026, powered by the Agentic Commerce Protocol (ACP) — an open standard co-developed with Stripe. Users can now discover and purchase products without leaving the conversation. Product results are organic and unsponsored. Etsy is live, over 1 million Shopify merchants are in the onboarding pipeline, and PayPal has adopted ACP to expand its merchant coverage. With an estimated 50 million shopping-related queries per day, ChatGPT is quickly becoming the largest AI-native commerce channel.
Google introduced agentic checkout in AI Mode in late 2025, enabling purchases directly within search results for eligible merchants. The feature works with merchants who support guest checkout and Google Pay — including retailers like Wayfair, Chewy, and select Shopify stores. At NRF 2026, Google also unveiled the Universal Commerce Protocol (UCP), co-developed with Shopify and backed by over 20 partners including Visa, Mastercard, Target, and Walmart.
Perplexity expanded its Buy with Pro shopping feature to all users in November 2025 through a PayPal partnership, connecting shoppers to 5,000+ merchants with in-chat checkout.
Amazon launched its Buy for Me feature, allowing users to purchase from third-party retailers without leaving the Amazon app. The scope has expanded from roughly 65,000 to over 500,000 products, though it remains limited to single-item orders and doesn’t carry Amazon’s A-to-Z guarantee.
These are real, working products — not demos. The progress over the past twelve months has been remarkable.
The Universality Gap
But there's a catch. Every approach listed above depends on merchant participation. OpenAI's ACP requires merchants to implement the protocol and provide a product feed. Google's agentic checkout requires guest checkout and Google Pay support. Perplexity's in-chat purchases work with its partner merchants. Amazon's Buy for Me covers a curated catalog.
That leaves millions of online merchants — independent stores on custom platforms, niche retailers, international sellers — out of reach for any agent.
This is the core infrastructure challenge of agentic commerce: the open web has millions of merchants, but the checkout systems that AI agents can programmatically complete cover only a fraction of them. Merchant fraud detection systems (bot protection, CAPTCHAs, device fingerprinting, velocity checks) were built to block exactly the kind of automated transactions that agents need to perform. Until checkout infrastructure can work across any merchant — not just those who've opted into a specific protocol — the promise of truly universal agentic commerce remains partially fulfilled.
Every major AI platform is racing to close the loop on shopping — moving from product recommendations to completed purchases inside the chat interface. Here's where the leading players stand as of early 2026.
OpenAI launched Instant Checkout in ChatGPT in February 2026, powered by the Agentic Commerce Protocol (ACP) — an open standard co-developed with Stripe. Users can now discover and purchase products without leaving the conversation. Product results are organic and unsponsored. Etsy is live, over 1 million Shopify merchants are in the onboarding pipeline, and PayPal has adopted ACP to expand its merchant coverage. With an estimated 50 million shopping-related queries per day, ChatGPT is quickly becoming the largest AI-native commerce channel.
Google introduced agentic checkout in AI Mode in late 2025, enabling purchases directly within search results for eligible merchants. The feature works with merchants who support guest checkout and Google Pay — including retailers like Wayfair, Chewy, and select Shopify stores. At NRF 2026, Google also unveiled the Universal Commerce Protocol (UCP), co-developed with Shopify and backed by over 20 partners including Visa, Mastercard, Target, and Walmart.
Perplexity expanded its Buy with Pro shopping feature to all users in November 2025 through a PayPal partnership, connecting shoppers to 5,000+ merchants with in-chat checkout.
Amazon launched its Buy for Me feature, allowing users to purchase from third-party retailers without leaving the Amazon app. The scope has expanded from roughly 65,000 to over 500,000 products, though it remains limited to single-item orders and doesn’t carry Amazon’s A-to-Z guarantee.
These are real, working products — not demos. The progress over the past twelve months has been remarkable.
The Universality Gap
But there's a catch. Every approach listed above depends on merchant participation. OpenAI's ACP requires merchants to implement the protocol and provide a product feed. Google's agentic checkout requires guest checkout and Google Pay support. Perplexity's in-chat purchases work with its partner merchants. Amazon's Buy for Me covers a curated catalog.
That leaves millions of online merchants — independent stores on custom platforms, niche retailers, international sellers — out of reach for any agent.
This is the core infrastructure challenge of agentic commerce: the open web has millions of merchants, but the checkout systems that AI agents can programmatically complete cover only a fraction of them. Merchant fraud detection systems (bot protection, CAPTCHAs, device fingerprinting, velocity checks) were built to block exactly the kind of automated transactions that agents need to perform. Until checkout infrastructure can work across any merchant — not just those who've opted into a specific protocol — the promise of truly universal agentic commerce remains partially fulfilled.
The Emerging Agentic Commerce Protocol Stack
Beyond individual platform implementations, a broader infrastructure layer is taking shape — a set of open protocols designed to standardize how agents, merchants, and payment providers interact.
Agentic Commerce Protocol (ACP), co-developed by OpenAI and Stripe under an Apache 2.0 license, defines the language for AI agents to programmatically discover products, initiate checkouts, and complete purchases. For merchants already on Stripe, enabling ACP can require as little as one line of code. For Shopify and Etsy merchants, it’s available out of the box.
Universal Commerce Protocol (UCP), developed by Google in collaboration with Shopify, Etsy, and over 20 partners, establishes a broader open standard covering the full shopping journey — from product discovery through fulfillment. UCP is designed to be platform-agnostic, working across Google surfaces, third-party agents, and any commerce interface that adopts it.
Agent Payments Protocol (AP2), also from Google, works alongside UCP to handle the payment layer specifically — integrating with providers like Mastercard, PayPal, and Visa to standardize how agents authorize and process transactions.
These protocols are complementary rather than competing. ACP focuses on the checkout transaction itself; UCP covers the broader commerce journey; AP2 handles payment authorization. Together, they're building the plumbing for a multi-agent commerce ecosystem.
The open question is coverage. These protocols work well for merchants on major platforms — Shopify alone accounts for millions of stores. But adoption takes time, and the merchants currently accessible through these protocols still represent a fraction of the total e-commerce landscape. For agents that need to transact across the full breadth of the open web, protocol adoption alone may not be enough.
Beyond individual platform implementations, a broader infrastructure layer is taking shape — a set of open protocols designed to standardize how agents, merchants, and payment providers interact.
Agentic Commerce Protocol (ACP), co-developed by OpenAI and Stripe under an Apache 2.0 license, defines the language for AI agents to programmatically discover products, initiate checkouts, and complete purchases. For merchants already on Stripe, enabling ACP can require as little as one line of code. For Shopify and Etsy merchants, it’s available out of the box.
Universal Commerce Protocol (UCP), developed by Google in collaboration with Shopify, Etsy, and over 20 partners, establishes a broader open standard covering the full shopping journey — from product discovery through fulfillment. UCP is designed to be platform-agnostic, working across Google surfaces, third-party agents, and any commerce interface that adopts it.
Agent Payments Protocol (AP2), also from Google, works alongside UCP to handle the payment layer specifically — integrating with providers like Mastercard, PayPal, and Visa to standardize how agents authorize and process transactions.
These protocols are complementary rather than competing. ACP focuses on the checkout transaction itself; UCP covers the broader commerce journey; AP2 handles payment authorization. Together, they're building the plumbing for a multi-agent commerce ecosystem.
The open question is coverage. These protocols work well for merchants on major platforms — Shopify alone accounts for millions of stores. But adoption takes time, and the merchants currently accessible through these protocols still represent a fraction of the total e-commerce landscape. For agents that need to transact across the full breadth of the open web, protocol adoption alone may not be enough.
Beyond individual platform implementations, a broader infrastructure layer is taking shape — a set of open protocols designed to standardize how agents, merchants, and payment providers interact.
Agentic Commerce Protocol (ACP), co-developed by OpenAI and Stripe under an Apache 2.0 license, defines the language for AI agents to programmatically discover products, initiate checkouts, and complete purchases. For merchants already on Stripe, enabling ACP can require as little as one line of code. For Shopify and Etsy merchants, it’s available out of the box.
Universal Commerce Protocol (UCP), developed by Google in collaboration with Shopify, Etsy, and over 20 partners, establishes a broader open standard covering the full shopping journey — from product discovery through fulfillment. UCP is designed to be platform-agnostic, working across Google surfaces, third-party agents, and any commerce interface that adopts it.
Agent Payments Protocol (AP2), also from Google, works alongside UCP to handle the payment layer specifically — integrating with providers like Mastercard, PayPal, and Visa to standardize how agents authorize and process transactions.
These protocols are complementary rather than competing. ACP focuses on the checkout transaction itself; UCP covers the broader commerce journey; AP2 handles payment authorization. Together, they're building the plumbing for a multi-agent commerce ecosystem.
The open question is coverage. These protocols work well for merchants on major platforms — Shopify alone accounts for millions of stores. But adoption takes time, and the merchants currently accessible through these protocols still represent a fraction of the total e-commerce landscape. For agents that need to transact across the full breadth of the open web, protocol adoption alone may not be enough.
Universal Checkout: The Missing Infrastructure Layer
There's a third approach — one that doesn't wait for merchants to opt in.
Rye built what it calls a Universal Checkout API: infrastructure that lets any AI agent purchase from any online merchant using nothing more than a product URL and a tokenized payment method. No merchant integration. No protocol adoption. No product feed.
The system works by combining browser-automation agents that can navigate arbitrary checkout flows, a caching layer that converts successful checkout traces into deterministic workflows (making repeat purchases up to 8x faster), and a fraud-mitigation proxy layer that prevents merchant systems from flagging legitimate agent transactions. The results are production-grade: Sub-35-second offer resolution across any store, and 99.9% order reliability with sub-10-second checkout latency on major merchant surfaces like Amazon and Shopify through direct integrations.
The contract is deliberately simple: product URL in, confirmed order out. The complexity — resolving prices, calculating taxes and shipping, handling payment securely, navigating each merchant's unique checkout flow — is abstracted away. Integrators never handle raw card data; the API uses tokenized payment methods, keeping PCI DSS scope off the developer's plate.
This matters because it means agentic commerce doesn't have to wait for universal merchant adoption of any single protocol. An agent using Rye's infrastructure can buy from a Shopify store, an Amazon listing, or an independent boutique running a custom checkout — all through the same API.
For developers building AI agents, commerce platforms, or any application where purchasing needs to happen programmatically, this is the difference between "works with some merchants" and "works with any merchant." You can explore the technical details in the Rye documentation.
There's a third approach — one that doesn't wait for merchants to opt in.
Rye built what it calls a Universal Checkout API: infrastructure that lets any AI agent purchase from any online merchant using nothing more than a product URL and a tokenized payment method. No merchant integration. No protocol adoption. No product feed.
The system works by combining browser-automation agents that can navigate arbitrary checkout flows, a caching layer that converts successful checkout traces into deterministic workflows (making repeat purchases up to 8x faster), and a fraud-mitigation proxy layer that prevents merchant systems from flagging legitimate agent transactions. The results are production-grade: Sub-35-second offer resolution across any store, and 99.9% order reliability with sub-10-second checkout latency on major merchant surfaces like Amazon and Shopify through direct integrations.
The contract is deliberately simple: product URL in, confirmed order out. The complexity — resolving prices, calculating taxes and shipping, handling payment securely, navigating each merchant's unique checkout flow — is abstracted away. Integrators never handle raw card data; the API uses tokenized payment methods, keeping PCI DSS scope off the developer's plate.
This matters because it means agentic commerce doesn't have to wait for universal merchant adoption of any single protocol. An agent using Rye's infrastructure can buy from a Shopify store, an Amazon listing, or an independent boutique running a custom checkout — all through the same API.
For developers building AI agents, commerce platforms, or any application where purchasing needs to happen programmatically, this is the difference between "works with some merchants" and "works with any merchant." You can explore the technical details in the Rye documentation.
There's a third approach — one that doesn't wait for merchants to opt in.
Rye built what it calls a Universal Checkout API: infrastructure that lets any AI agent purchase from any online merchant using nothing more than a product URL and a tokenized payment method. No merchant integration. No protocol adoption. No product feed.
The system works by combining browser-automation agents that can navigate arbitrary checkout flows, a caching layer that converts successful checkout traces into deterministic workflows (making repeat purchases up to 8x faster), and a fraud-mitigation proxy layer that prevents merchant systems from flagging legitimate agent transactions. The results are production-grade: Sub-35-second offer resolution across any store, and 99.9% order reliability with sub-10-second checkout latency on major merchant surfaces like Amazon and Shopify through direct integrations.
The contract is deliberately simple: product URL in, confirmed order out. The complexity — resolving prices, calculating taxes and shipping, handling payment securely, navigating each merchant's unique checkout flow — is abstracted away. Integrators never handle raw card data; the API uses tokenized payment methods, keeping PCI DSS scope off the developer's plate.
This matters because it means agentic commerce doesn't have to wait for universal merchant adoption of any single protocol. An agent using Rye's infrastructure can buy from a Shopify store, an Amazon listing, or an independent boutique running a custom checkout — all through the same API.
For developers building AI agents, commerce platforms, or any application where purchasing needs to happen programmatically, this is the difference between "works with some merchants" and "works with any merchant." You can explore the technical details in the Rye documentation.
Agentic Commerce Use Cases and Examples
Agentic commerce isn't theoretical. Real companies are deploying agent-driven purchasing workflows right now, across industries you might not expect.
B2B procurement. Logistics companies are using agents to auto-purchase shipping supplies when inventory runs low. The agent monitors stock levels, compares vendor prices, and places orders — no purchase order approval chain required for transactions under a set threshold.
Marketing operations. Teams are deploying agents that monitor ad campaign performance and automatically top up advertising credits when spend hits a threshold, optimizing for budget efficiency without manual intervention.
Facilities management. Recurring supply orders — cleaning products, office supplies, printer ink — are being delegated to agents that optimize for price and delivery timing across multiple vendors.
Gifting and personal shopping. Consumer-facing applications are letting users describe what they want ("a birthday gift for my mom who likes gardening, under $50") and having agents research, select, and purchase a curated option.
Healthcare supply chain. Procurement teams are exploring agent-driven purchasing for medical supplies, where price volatility and supply constraints make automated comparison shopping particularly valuable.
The pattern across all of these: high-frequency, repeatable purchases where the decision criteria can be clearly defined and the cost of human labor to execute them is disproportionate to the transaction value.
Agentic commerce isn't theoretical. Real companies are deploying agent-driven purchasing workflows right now, across industries you might not expect.
B2B procurement. Logistics companies are using agents to auto-purchase shipping supplies when inventory runs low. The agent monitors stock levels, compares vendor prices, and places orders — no purchase order approval chain required for transactions under a set threshold.
Marketing operations. Teams are deploying agents that monitor ad campaign performance and automatically top up advertising credits when spend hits a threshold, optimizing for budget efficiency without manual intervention.
Facilities management. Recurring supply orders — cleaning products, office supplies, printer ink — are being delegated to agents that optimize for price and delivery timing across multiple vendors.
Gifting and personal shopping. Consumer-facing applications are letting users describe what they want ("a birthday gift for my mom who likes gardening, under $50") and having agents research, select, and purchase a curated option.
Healthcare supply chain. Procurement teams are exploring agent-driven purchasing for medical supplies, where price volatility and supply constraints make automated comparison shopping particularly valuable.
The pattern across all of these: high-frequency, repeatable purchases where the decision criteria can be clearly defined and the cost of human labor to execute them is disproportionate to the transaction value.
Agentic commerce isn't theoretical. Real companies are deploying agent-driven purchasing workflows right now, across industries you might not expect.
B2B procurement. Logistics companies are using agents to auto-purchase shipping supplies when inventory runs low. The agent monitors stock levels, compares vendor prices, and places orders — no purchase order approval chain required for transactions under a set threshold.
Marketing operations. Teams are deploying agents that monitor ad campaign performance and automatically top up advertising credits when spend hits a threshold, optimizing for budget efficiency without manual intervention.
Facilities management. Recurring supply orders — cleaning products, office supplies, printer ink — are being delegated to agents that optimize for price and delivery timing across multiple vendors.
Gifting and personal shopping. Consumer-facing applications are letting users describe what they want ("a birthday gift for my mom who likes gardening, under $50") and having agents research, select, and purchase a curated option.
Healthcare supply chain. Procurement teams are exploring agent-driven purchasing for medical supplies, where price volatility and supply constraints make automated comparison shopping particularly valuable.
The pattern across all of these: high-frequency, repeatable purchases where the decision criteria can be clearly defined and the cost of human labor to execute them is disproportionate to the transaction value.
Agentic Commerce Market Size and Growth
The data points tell a clear story about where this is heading.
According to Edgar, Dunn & Company, the total addressable market for agentic commerce is projected to reach $135 billion in 2025 and grow to $1.7 trillion by 2030 — a compound annual growth rate of 67%.
McKinsey’s research suggests AI agents could mediate $3 trillion to $5 trillion of global consumer commerce by 2030 under moderate scenarios. Morgan Stanley estimates agent-driven spending could capture 10–20% of the total U.S. online retail market by the same year.
On the consumer side, adoption is accelerating faster than most predicted. An IBM/NRF study found that 45% of consumers already use AI in some part of their buying journey. Shopping-related generative AI searches grew 4,700% between July 2024 and July 2025.
And perhaps the most telling stat: roughly 23% of Americans have already made a purchase via AI in the past month, according to Morgan Stanley — and adoption is accelerating across demographics. Trust is building fast.
The data points tell a clear story about where this is heading.
According to Edgar, Dunn & Company, the total addressable market for agentic commerce is projected to reach $135 billion in 2025 and grow to $1.7 trillion by 2030 — a compound annual growth rate of 67%.
McKinsey’s research suggests AI agents could mediate $3 trillion to $5 trillion of global consumer commerce by 2030 under moderate scenarios. Morgan Stanley estimates agent-driven spending could capture 10–20% of the total U.S. online retail market by the same year.
On the consumer side, adoption is accelerating faster than most predicted. An IBM/NRF study found that 45% of consumers already use AI in some part of their buying journey. Shopping-related generative AI searches grew 4,700% between July 2024 and July 2025.
And perhaps the most telling stat: roughly 23% of Americans have already made a purchase via AI in the past month, according to Morgan Stanley — and adoption is accelerating across demographics. Trust is building fast.
The data points tell a clear story about where this is heading.
According to Edgar, Dunn & Company, the total addressable market for agentic commerce is projected to reach $135 billion in 2025 and grow to $1.7 trillion by 2030 — a compound annual growth rate of 67%.
McKinsey’s research suggests AI agents could mediate $3 trillion to $5 trillion of global consumer commerce by 2030 under moderate scenarios. Morgan Stanley estimates agent-driven spending could capture 10–20% of the total U.S. online retail market by the same year.
On the consumer side, adoption is accelerating faster than most predicted. An IBM/NRF study found that 45% of consumers already use AI in some part of their buying journey. Shopping-related generative AI searches grew 4,700% between July 2024 and July 2025.
And perhaps the most telling stat: roughly 23% of Americans have already made a purchase via AI in the past month, according to Morgan Stanley — and adoption is accelerating across demographics. Trust is building fast.
The Future of Agentic Commerce: What Needs to Happen Next
Agentic commerce is past the proof-of-concept stage but still early in the adoption curve. Several things need to mature before it becomes the default way goods are bought and sold.
Checkout infrastructure needs to become truly universal. Protocols like ACP and UCP are important steps, but the long tail of e-commerce — millions of merchants on custom platforms — needs to be reachable too. Solutions like Rye's Universal Checkout API that work without merchant integration will be critical for bridging that gap.
Consumer trust needs to keep building. People are warming up to AI-assisted purchasing, but the industry needs to deliver on reliability and transparency. Agents that make bad purchases or can't handle returns will set the entire category back.
Payment security needs to evolve. Tokenization, programmable cards, and agent-specific payment rails need to become standard — not experimental. The PCI compliance burden for developers building agentic applications needs to shrink, not grow. Companies like Basis Theory, Nekuda, Skyfire, and Prava are building alongside infrastructure providers like Rye to make agent-specific payment and identity rails production-ready.
Discovery needs to go beyond product feeds. Today's agent shopping experiences are limited to whatever products are in a structured catalog. The next generation of agents will need to discover and evaluate products across the open web — the same way humans do, but at machine speed.
Agentic commerce is past the proof-of-concept stage but still early in the adoption curve. Several things need to mature before it becomes the default way goods are bought and sold.
Checkout infrastructure needs to become truly universal. Protocols like ACP and UCP are important steps, but the long tail of e-commerce — millions of merchants on custom platforms — needs to be reachable too. Solutions like Rye's Universal Checkout API that work without merchant integration will be critical for bridging that gap.
Consumer trust needs to keep building. People are warming up to AI-assisted purchasing, but the industry needs to deliver on reliability and transparency. Agents that make bad purchases or can't handle returns will set the entire category back.
Payment security needs to evolve. Tokenization, programmable cards, and agent-specific payment rails need to become standard — not experimental. The PCI compliance burden for developers building agentic applications needs to shrink, not grow. Companies like Basis Theory, Nekuda, Skyfire, and Prava are building alongside infrastructure providers like Rye to make agent-specific payment and identity rails production-ready.
Discovery needs to go beyond product feeds. Today's agent shopping experiences are limited to whatever products are in a structured catalog. The next generation of agents will need to discover and evaluate products across the open web — the same way humans do, but at machine speed.
Agentic commerce is past the proof-of-concept stage but still early in the adoption curve. Several things need to mature before it becomes the default way goods are bought and sold.
Checkout infrastructure needs to become truly universal. Protocols like ACP and UCP are important steps, but the long tail of e-commerce — millions of merchants on custom platforms — needs to be reachable too. Solutions like Rye's Universal Checkout API that work without merchant integration will be critical for bridging that gap.
Consumer trust needs to keep building. People are warming up to AI-assisted purchasing, but the industry needs to deliver on reliability and transparency. Agents that make bad purchases or can't handle returns will set the entire category back.
Payment security needs to evolve. Tokenization, programmable cards, and agent-specific payment rails need to become standard — not experimental. The PCI compliance burden for developers building agentic applications needs to shrink, not grow. Companies like Basis Theory, Nekuda, Skyfire, and Prava are building alongside infrastructure providers like Rye to make agent-specific payment and identity rails production-ready.
Discovery needs to go beyond product feeds. Today's agent shopping experiences are limited to whatever products are in a structured catalog. The next generation of agents will need to discover and evaluate products across the open web — the same way humans do, but at machine speed.
Frequently Asked Questions
What is the difference between agentic commerce and online shopping?
Traditional online shopping is human-directed at every step — you search, you browse, you compare, you checkout. Agentic commerce delegates some or all of those steps to an AI agent that can act autonomously within guardrails you define. The agent handles research, price comparison, offer resolution, and purchase execution. You set the criteria (budget, preferences, timing); the agent handles the workflow.
Is agentic commerce safe?
Safety depends on the infrastructure layer. Production-grade agentic commerce systems use tokenized payment methods so AI agents never see raw card numbers, spending limits and approval thresholds so agents can't exceed authorized budgets, audit trails that log every action the agent takes, and established buyer-protection policies from payment providers like PayPal, Visa, and Stripe. Consumer trust is growing quickly — roughly 23% of Americans have already made a purchase via AI in the past month, according to Morgan Stanley, and adoption is accelerating across demographics.
How is agentic commerce different from chatbot shopping?
Chatbots are conversational interfaces that help you find products, but they still require you to confirm and complete the purchase. They reduce friction at the discovery stage. Agentic commerce systems can close the entire loop — from intent to confirmed order — without requiring human intervention at each step. The distinction is autonomy: a chatbot assists; an agent executes.
What are the main agentic commerce protocols?
Three protocols are shaping the space. OpenAI’s Agentic Commerce Protocol (ACP), co-developed with Stripe, standardizes how agents discover and purchase products — it’s live in ChatGPT with Etsy and over 1 million Shopify merchants onboarding. Google’s Universal Commerce Protocol (UCP) covers the full shopping journey from discovery to fulfillment, backed by 20+ partners including Visa, Mastercard, and Walmart. Google’s Agent Payments Protocol (AP2) handles the payment authorization layer specifically, integrating with major card networks and payment providers.
Can AI agents buy from any online store?
Not yet — but it's getting closer. Protocol-based approaches (ACP, UCP) require merchants to opt in, which limits coverage to participating stores. Universal checkout infrastructure like Rye's API takes a different approach, using browser automation and fraud mitigation to complete purchases on any merchant site using just a product URL — no merchant integration required. The combination of protocol adoption and universal checkout infrastructure is what will eventually make any-merchant purchasing a reality.
What is the difference between agentic commerce and online shopping?
Traditional online shopping is human-directed at every step — you search, you browse, you compare, you checkout. Agentic commerce delegates some or all of those steps to an AI agent that can act autonomously within guardrails you define. The agent handles research, price comparison, offer resolution, and purchase execution. You set the criteria (budget, preferences, timing); the agent handles the workflow.
Is agentic commerce safe?
Safety depends on the infrastructure layer. Production-grade agentic commerce systems use tokenized payment methods so AI agents never see raw card numbers, spending limits and approval thresholds so agents can't exceed authorized budgets, audit trails that log every action the agent takes, and established buyer-protection policies from payment providers like PayPal, Visa, and Stripe. Consumer trust is growing quickly — roughly 23% of Americans have already made a purchase via AI in the past month, according to Morgan Stanley, and adoption is accelerating across demographics.
How is agentic commerce different from chatbot shopping?
Chatbots are conversational interfaces that help you find products, but they still require you to confirm and complete the purchase. They reduce friction at the discovery stage. Agentic commerce systems can close the entire loop — from intent to confirmed order — without requiring human intervention at each step. The distinction is autonomy: a chatbot assists; an agent executes.
What are the main agentic commerce protocols?
Three protocols are shaping the space. OpenAI’s Agentic Commerce Protocol (ACP), co-developed with Stripe, standardizes how agents discover and purchase products — it’s live in ChatGPT with Etsy and over 1 million Shopify merchants onboarding. Google’s Universal Commerce Protocol (UCP) covers the full shopping journey from discovery to fulfillment, backed by 20+ partners including Visa, Mastercard, and Walmart. Google’s Agent Payments Protocol (AP2) handles the payment authorization layer specifically, integrating with major card networks and payment providers.
Can AI agents buy from any online store?
Not yet — but it's getting closer. Protocol-based approaches (ACP, UCP) require merchants to opt in, which limits coverage to participating stores. Universal checkout infrastructure like Rye's API takes a different approach, using browser automation and fraud mitigation to complete purchases on any merchant site using just a product URL — no merchant integration required. The combination of protocol adoption and universal checkout infrastructure is what will eventually make any-merchant purchasing a reality.
What is the difference between agentic commerce and online shopping?
Traditional online shopping is human-directed at every step — you search, you browse, you compare, you checkout. Agentic commerce delegates some or all of those steps to an AI agent that can act autonomously within guardrails you define. The agent handles research, price comparison, offer resolution, and purchase execution. You set the criteria (budget, preferences, timing); the agent handles the workflow.
Is agentic commerce safe?
Safety depends on the infrastructure layer. Production-grade agentic commerce systems use tokenized payment methods so AI agents never see raw card numbers, spending limits and approval thresholds so agents can't exceed authorized budgets, audit trails that log every action the agent takes, and established buyer-protection policies from payment providers like PayPal, Visa, and Stripe. Consumer trust is growing quickly — roughly 23% of Americans have already made a purchase via AI in the past month, according to Morgan Stanley, and adoption is accelerating across demographics.
How is agentic commerce different from chatbot shopping?
Chatbots are conversational interfaces that help you find products, but they still require you to confirm and complete the purchase. They reduce friction at the discovery stage. Agentic commerce systems can close the entire loop — from intent to confirmed order — without requiring human intervention at each step. The distinction is autonomy: a chatbot assists; an agent executes.
What are the main agentic commerce protocols?
Three protocols are shaping the space. OpenAI’s Agentic Commerce Protocol (ACP), co-developed with Stripe, standardizes how agents discover and purchase products — it’s live in ChatGPT with Etsy and over 1 million Shopify merchants onboarding. Google’s Universal Commerce Protocol (UCP) covers the full shopping journey from discovery to fulfillment, backed by 20+ partners including Visa, Mastercard, and Walmart. Google’s Agent Payments Protocol (AP2) handles the payment authorization layer specifically, integrating with major card networks and payment providers.
Can AI agents buy from any online store?
Not yet — but it's getting closer. Protocol-based approaches (ACP, UCP) require merchants to opt in, which limits coverage to participating stores. Universal checkout infrastructure like Rye's API takes a different approach, using browser automation and fraud mitigation to complete purchases on any merchant site using just a product URL — no merchant integration required. The combination of protocol adoption and universal checkout infrastructure is what will eventually make any-merchant purchasing a reality.
The Bottom Line
Agentic commerce is the next logical step in how the internet handles transactions. We went from browsing catalogs to searching for products to getting personalized recommendations. The next step — having an intelligent agent handle the entire purchase on your behalf — was always inevitable. What changed is that the AI, the infrastructure, and the payment systems have finally caught up.
The companies that will win this transition aren't just the ones building the smartest agents. They're the ones building the infrastructure that makes any agent capable of buying from any merchant, reliably, at scale.
If you're building AI agents or commerce applications, the checkout problem is the one worth solving first. Everything else — discovery, recommendations, personalization — matters less if your agent can't actually complete the purchase.
Agentic commerce is the next logical step in how the internet handles transactions. We went from browsing catalogs to searching for products to getting personalized recommendations. The next step — having an intelligent agent handle the entire purchase on your behalf — was always inevitable. What changed is that the AI, the infrastructure, and the payment systems have finally caught up.
The companies that will win this transition aren't just the ones building the smartest agents. They're the ones building the infrastructure that makes any agent capable of buying from any merchant, reliably, at scale.
If you're building AI agents or commerce applications, the checkout problem is the one worth solving first. Everything else — discovery, recommendations, personalization — matters less if your agent can't actually complete the purchase.
Agentic commerce is the next logical step in how the internet handles transactions. We went from browsing catalogs to searching for products to getting personalized recommendations. The next step — having an intelligent agent handle the entire purchase on your behalf — was always inevitable. What changed is that the AI, the infrastructure, and the payment systems have finally caught up.
The companies that will win this transition aren't just the ones building the smartest agents. They're the ones building the infrastructure that makes any agent capable of buying from any merchant, reliably, at scale.
If you're building AI agents or commerce applications, the checkout problem is the one worth solving first. Everything else — discovery, recommendations, personalization — matters less if your agent can't actually complete the purchase.
Start Building
Rye's Universal Checkout API lets your AI agent purchase from any merchant — no merchant integration required. Get your API key and start building in minutes.
Read the docs → docs.rye.com/quick-start
Rye's Universal Checkout API lets your AI agent purchase from any merchant — no merchant integration required. Get your API key and start building in minutes.
Read the docs → docs.rye.com/quick-start
Rye's Universal Checkout API lets your AI agent purchase from any merchant — no merchant integration required. Get your API key and start building in minutes.
Read the docs → docs.rye.com/quick-start
Related Articles
Related Articles
Stop the redirect.
Start the revenue.
Stop the redirect.
Start the revenue.
Stop the redirect.
Start the revenue.
Product
Resources