Table of Contents
The shipping and tax problem: Why agents struggle to close the sale
Sophia Willows
Head of Engineering @ Rye
Aug 27, 2025
Dynamic shipping and tax break most agent checkouts. Rye's Universal Checkout API surfaces true totals fast, so users can confirm their order without growing impatient.
One of the toughest challenges in agentic commerce is showing not the final cost, including shipping and tax. Rye has solved this issue with its approach to universal checkout.
One of the toughest challenges in agentic commerce is showing not the final cost, including shipping and tax. Rye has solved this issue with its approach to universal checkout.
One of the toughest challenges in agentic commerce is showing not the final cost, including shipping and tax. Rye has solved this issue with its approach to universal checkout.
Why agents have a hard time with tax and shipping
Shipping and tax must be calculated for each order, and can only be calculated by the merchant. Unlike SKU prices, which can be crawled and cached, these order-specific charges are based on dynamic variables like shipping distance, carrier availability, service tiers, and jurisdiction-specific taxes.
To get real-time information from an arbitrary merchant, an agent must be able to read its webpage. But most get tripped up, either by explicit traps like anti-fraud filters, or confusion over DOM elements like unexpected popups.
Most other so-called agentic platforms fake it to finish the sale. Some rely on manual fulfillment teams to retrieve tax and shipping costs when automation fails, which gets the order through but keeps users waiting for a final price for several minutes. Others avoid the problem altogether by showing only the SKU price and absorbing tax and shipping. Instant (and cheaper!) for the user, but ruinously expensive at scale. Neither approach works as production infrastructure.
How Rye makes automated checkout work
Rye solved shipping and tax by solving automated checkout end-to-end, with an agent that can successfully and repeatably navigate the long tail of e-commerce sites.
The system runs in two phases. On the first pass, an LLM navigates a merchant’s checkout flow in exploratory mode. A filter simplifies the DOM by suppressing popups, banners, and irrelevant fields so the model can focus on what matters. Guardrails catch obvious errors. This exploratory run produces a trace of how to reach the point in checkout where the merchant reveals shipping options and tax.
On the second pass, Rye converts that trace into a deterministic workflow. On subsequent orders, the workflow executes step by step without invoking the LLM. If the site changes, only the failing step is rerun with an LLM, after which the workflow self-heals. Because the LLM is limited to exploration and occasional healing, Rye can rely on smaller, faster models. Latency is no longer dictated by expensive reasoning cycles.
None of these techniques would matter if we couldn’t reliably get to the checkout page. Rye’s Merchant Risk Adapter for fraud-safe navigation leverages residential IPs, geo-proximal sessions, and human-like pacing to appear as a real shopper to merchants.
Rye solved shipping and tax by solving automated checkout end-to-end, with an agent that can successfully and repeatably navigate the long tail of e-commerce sites.
The system runs in two phases. On the first pass, an LLM navigates a merchant’s checkout flow in exploratory mode. A filter simplifies the DOM by suppressing popups, banners, and irrelevant fields so the model can focus on what matters. Guardrails catch obvious errors. This exploratory run produces a trace of how to reach the point in checkout where the merchant reveals shipping options and tax.
On the second pass, Rye converts that trace into a deterministic workflow. On subsequent orders, the workflow executes step by step without invoking the LLM. If the site changes, only the failing step is rerun with an LLM, after which the workflow self-heals. Because the LLM is limited to exploration and occasional healing, Rye can rely on smaller, faster models. Latency is no longer dictated by expensive reasoning cycles.
None of these techniques would matter if we couldn’t reliably get to the checkout page. Rye’s Merchant Risk Adapter for fraud-safe navigation leverages residential IPs, geo-proximal sessions, and human-like pacing to appear as a real shopper to merchants.
Rye solved shipping and tax by solving automated checkout end-to-end, with an agent that can successfully and repeatably navigate the long tail of e-commerce sites.
The system runs in two phases. On the first pass, an LLM navigates a merchant’s checkout flow in exploratory mode. A filter simplifies the DOM by suppressing popups, banners, and irrelevant fields so the model can focus on what matters. Guardrails catch obvious errors. This exploratory run produces a trace of how to reach the point in checkout where the merchant reveals shipping options and tax.
On the second pass, Rye converts that trace into a deterministic workflow. On subsequent orders, the workflow executes step by step without invoking the LLM. If the site changes, only the failing step is rerun with an LLM, after which the workflow self-heals. Because the LLM is limited to exploration and occasional healing, Rye can rely on smaller, faster models. Latency is no longer dictated by expensive reasoning cycles.
None of these techniques would matter if we couldn’t reliably get to the checkout page. Rye’s Merchant Risk Adapter for fraud-safe navigation leverages residential IPs, geo-proximal sessions, and human-like pacing to appear as a real shopper to merchants.
Performance targets
Rye measures performance in two ways: offer resolution and checkout.
Offer resolution is the time it takes to surface a true landed cost—item price plus real-time shipping options and taxes—so the shopper can make a decision. This is a crucial metric for us, because every second a shopper has to wait increases their likelihood of abandoning the transaction.
By mid-September, we expect our browser-based flows across the open web to hit a mean of 35-second offer resolution for 90% of orders. This is dramatically faster than the minutes-long offer resolution seen from other agents. We also expect a ten-second checkout by then; we can do it so quickly because we use the same DOM session to complete the order as we used to generate it.
Note that we go a different route with Amazon and Shopify, plugging directly into their systems to deliver 5- and 12-second checkout, respectively. The high volumes of these platforms justify the custom builds for even faster speeds.
Rye measures performance in two ways: offer resolution and checkout.
Offer resolution is the time it takes to surface a true landed cost—item price plus real-time shipping options and taxes—so the shopper can make a decision. This is a crucial metric for us, because every second a shopper has to wait increases their likelihood of abandoning the transaction.
By mid-September, we expect our browser-based flows across the open web to hit a mean of 35-second offer resolution for 90% of orders. This is dramatically faster than the minutes-long offer resolution seen from other agents. We also expect a ten-second checkout by then; we can do it so quickly because we use the same DOM session to complete the order as we used to generate it.
Note that we go a different route with Amazon and Shopify, plugging directly into their systems to deliver 5- and 12-second checkout, respectively. The high volumes of these platforms justify the custom builds for even faster speeds.
Rye measures performance in two ways: offer resolution and checkout.
Offer resolution is the time it takes to surface a true landed cost—item price plus real-time shipping options and taxes—so the shopper can make a decision. This is a crucial metric for us, because every second a shopper has to wait increases their likelihood of abandoning the transaction.
By mid-September, we expect our browser-based flows across the open web to hit a mean of 35-second offer resolution for 90% of orders. This is dramatically faster than the minutes-long offer resolution seen from other agents. We also expect a ten-second checkout by then; we can do it so quickly because we use the same DOM session to complete the order as we used to generate it.
Note that we go a different route with Amazon and Shopify, plugging directly into their systems to deliver 5- and 12-second checkout, respectively. The high volumes of these platforms justify the custom builds for even faster speeds.
Why it matters
With Rye, the number the shopper sees is the number they pay. For integrators, Rye’s approach means no hidden subsidies, rare human fallback, and confident end-to-end demos.
By transforming exploratory agent runs into prescriptive workflows, Rye makes agentic checkout production-ready. Shipping and tax stop being the breaking point and become just another step in a fast, reliable system for closing the sale. See more in the docs.
With Rye, the number the shopper sees is the number they pay. For integrators, Rye’s approach means no hidden subsidies, rare human fallback, and confident end-to-end demos.
By transforming exploratory agent runs into prescriptive workflows, Rye makes agentic checkout production-ready. Shipping and tax stop being the breaking point and become just another step in a fast, reliable system for closing the sale. See more in the docs.
With Rye, the number the shopper sees is the number they pay. For integrators, Rye’s approach means no hidden subsidies, rare human fallback, and confident end-to-end demos.
By transforming exploratory agent runs into prescriptive workflows, Rye makes agentic checkout production-ready. Shipping and tax stop being the breaking point and become just another step in a fast, reliable system for closing the sale. See more in the docs.
Related articles
Related articles
Monetize
your AI platform
with shopping.
Monetize
your AI platform
with shopping.
Monetize
your AI platform
with shopping.
Product
Resources