Last September, Release Notes 231 introduced the Icecat MCP Server in Beta. At the time, its functionality was straightforward: when an AI agent supplied a product code and brand name, the server could return the corresponding product information.
Since then, the MCP Server has expanded considerably. It now offers nineteen tools and is listed in the official public MCP Registry. More importantly, after almost a year of real-world usage, we now have enough data to answer one of the questions we hear most often: how are partners actually using it?
We reviewed usage statistics on an account-by-account basis and spoke with the partners behind those integrations over several months. The results reveal five distinct use cases, including several we did not initially anticipate, and provide a clearer picture of how the Icecat MCP Server is being used in practice.
Monthly requests to the MCP server, September 2025 through 15 August 2026.
Five observations stand out from the data.
Growth happens in spikes rather than along a smooth curve. Each major increase corresponds to a partner activating an integration rather than a gradual rise in demand. July 2026 was our largest month, with just over 17,000 requests, five times the previous record. Around 59% of those requests came from a single account. Some spikes later declined as individual projects reached completion.
The number of active companies has grown steadily. We moved from four active companies in September 2025 to between 20 and 30 per month throughout the summer of 2026. We pay particular attention to this metric because it reflects how many partners are actively using the server, rather than letting a single high-volume integration dominate the overall picture.
Weekly engagement has roughly tripled. Looking specifically at companies making meaningful use of the server during a given week, activity has increased from around four or five companies per week in spring to between 10 and 15 per week across July and August.
The early months were deliberately constrained. In Release 232, we temporarily limited MCP access to Open Icecat data while studying usage patterns. The lower volumes recorded during autumn therefore reflect a deliberately narrower product rather than a lack of interest.
New integrations can scale quickly. The August 2026 data covers only the first 15 days and is tracking at a rate similar to July. However, the activity comes from an entirely different partner, which increased from approximately 70 calls on one day to more than 1,600 just two days later. Once a workflow proves effective, usage can scale rapidly.
One of our highest-volume partners is a retailer whose local language is not currently available as an Icecat locale. The retailer retrieves the complete product overview from MCP and independently translates the content as part of an automated workflow.
This was not a use case we originally designed for, but it demonstrates an interesting capability of the MCP Server. It can serve as a factual grounding source for translation pipelines because the model translates information derived from structured product data rather than scraped marketing copy.
It is worth adding, for anyone considering the same approach, that Icecat already offers a dedicated localization solution for partners who need high-quality translated product content. That approach is already adopted by multiple partners and does not require you to build a translation pipeline yourself. If that is of interest to your catalog, speak to your Icecat account manager or contact us via our contact form.
Several partners use the MCP Server for a deliberately straightforward purpose: retrieving product content for their own catalogs, much as they would through the Icecat API.
Although less experimental than some other applications, this may be the most broadly applicable use case.
MCP provides an alternative interface to the same product data. If a team already works with an AI assistant, there is no need to develop and maintain an API client simply to bring Icecat product content into its workflow. Instead, the assistant can connect to the server and retrieve the required information directly.
Specifications, images, descriptions, and identifiers are all returned as structured data. For teams without dedicated development capacity, this significantly lowers the barrier to working with Icecat content.
One advertising agency used the MCP Server intensively during two concentrated periods before its activity stopped. We now recognize this usage pattern as characteristic of a client project rather than an ongoing process.
The agency used real product specifications retrieved via MCP to generate optimized, unique content for paid search campaigns.
A second retailer follows a different approach. It uses MCP to generate review-style content based on a product’s genuine strengths and weaknesses, derived from its actual specifications. The resulting unique content is then used to support organic search traffic.
Although the applications differ, both demonstrate the same principle: specifications provide facts, and facts give generative models a reliable basis for producing accurate content.
MCP provides a convenient way to put those facts directly into the model’s workflow.
Many Icecat partners sell through marketplaces, each with its own publication requirements.
Amazon is an obvious example, but the same applies to Shopee, Lazada, and other platforms. Requirements for title length, bullet formats, mandatory attributes, and prohibited wording differ between marketplaces. They can also differ considerably from the standardized product content that partners maintain within their own systems.
Adapting content to every platform has traditionally required manual work. AI is well suited to reshaping content, but only when it begins with reliable facts.
If a model is asked to rewrite an existing marketplace listing, it may confidently introduce specifications that were never accurate in the first place. Its source is someone else’s marketing copy rather than the underlying product information.
This is where MCP provides a stronger foundation. It provides verified product information, including accurate specifications, identifiers, and images. An AI workflow can then transform standardized product content to meet each marketplace’s requirements without inventing information along the way.
The sequence is simple: retrieve the facts first, then reshape them according to the platform’s rules.
The same principle extends beyond marketplaces. Generative content pipelines need a factual anchor, and structured product data can provide it. Without that foundation, content generation can quickly become content invention.
Lena, Icecat’s AI Assistant, launched on Icecat.biz in Release 254, is also an MCP consumer, querying the server to answer customer questions about products.
Lena was developed for Icecat’s own support operations rather than as a demonstration of the MCP Server. However, she has since become one of the most informative accounts to observe.
She uses twelve different MCP tools, more than almost any partner. Her usage pattern also differs considerably from most other accounts. Rather than relying primarily on the general product overview, Lena makes greater use of image galleries and product identifiers, as these are often the resources needed in real customer-support conversations.
Perhaps the most notable characteristic is consistency.
Lena has been active almost every day since launch. This illustrates an important difference between an agent integration and a human-driven workflow: once deployed, the agent continues to operate. It does not forget that the integration exists or become occupied with another project.
Product search. Every tool currently available assumes that the user or agent already knows which product it wants to retrieve. Search would allow an agent to discover a product it cannot yet name. This remains the largest gap in the current functionality and is now under development.
Multilingual output through MCP. Icecat partners operate across multiple languages and markets, and we are actively improving how the server handles locales in its responses.
Simpler connection. Connecting an AI client to the MCP Server currently requires manually editing a configuration file. By comparison, many other connectors can be installed in only a few clicks through a standard authorization flow. Simplifying this first connection is therefore a priority. The faster a workflow can be established, the sooner it can become part of a partner’s regular operations.
A year ago, the Icecat MCP Server could answer one type of product query. Today, it supports translation workflows, straightforward content retrieval, SEO content generation, marketplace publishing, and AI-powered support assistants.
One of the most interesting observations from almost a year of usage is that the most consistent consumer of Icecat product data through the server is not a person. It is an AI.
If this pattern continues, and we increasingly believe it will, a meaningful share of product-content consumption could shift from humans browsing interfaces to agents querying data directly.
Agents do not need an attractive interface. What they need is product information that is complete, structured, consistent, and predictable. Conveniently, that is precisely what Icecat has spent the past twenty years building.
If you use the Icecat MCP Server, we would like to hear how you use it. Identifying these use cases required months of conversations with partners, and each one revealed insights that usage statistics alone could never provide.
Get in touch with your Icecat account manager or reply to this post and tell us what you are building.
Read further: Icecat, e-commerce, ecommerce, Icecat, MCP, product content
