Full Icecat Added to the MCP Server

After our September announcement, we have added Full Icecat support directly to our Model Context Protocol (MCP) server. This enhancement unlocks seamless access to complete catalog depth for enterprise workflows and AI agents everywhere for users that have a paid Full Icecat subscription.

Open and Full Icecat Support

In the pilot phase, only Open Icecat data was available. This phase allowed us to monitor adoption, user requirements, stability and security of this new agentic interface. The number of monthly active users (MAU) has been growing steadily to around 30 in July and August. Although the initial adoption started slower than we expected, given the huge attention for AI in the media, the upward trend in the first twelve months is sufficiently encouraging to do the next step. Making the MCP behavior compatible with our standard APIs, taking account user subscriptions and authorizations.

The decision directly addresses how the user community is built. Currently, about half of all active MCP server users hold Full Icecat subscriptions, while the other half rely on the free, open-source content subscription provided through free Open Icecat. Bridging this gap is therefore a key priority. By bringing Full Icecat capabilities into the MCP environment, high-volume users and enterprise teams are now able to stream millions of complete datasheets, deep specifications, and advanced asset feeds directly into their AI systems. See the manual for using our MCP server as a tool in your AI agent.

MCP Use Cases

In our usage analysis, we identified five major use cases: grounding translation workflows, MCP as a simple API, SEO/GEO content and review generation, meeting marketplace content requirements and AI Agent support. The variation in use cases is much higher already than we expected beforehand. Therefore, we can only assume that MCP support will lead to exploring even more application areas in the next period.

On the short term, we don’t forsee support to other agentic protocols like UCP. We strongly believe that MCP is sufficiently versatile in this phase of AI in commerce. When more data points are available on the adoption of various protocols we’ll reconsider our stance.

Development is moving quickly. Additional documentation, roll-out specifics, and technical setup guides will follow as the sprint progresses toward completion. Such new developments will also include MCP search capabilities and further powering our own AI agent Lena. Stay tuned for further details in the coming weeks.