In short: Shopware 6.7.10.0 (released May 6, 2026) introduces a new, experimental sales channel type for agentic commerce: products can be exposed as a JSONL feed for AI platforms, starting with the OpenAI integration. Orders and traffic from AI platforms are tracked through the existing affiliate infrastructure. It sounds like a feature checkbox - in reality it is the entry point into a channel where AI agents read your product data instead of humans clicking through your storefront.
What the new sales channel actually does
The Agentic Commerce sales channel is a dedicated channel type in Shopware that serves as the central entry point for AI-driven product distribution. The building blocks:
- JSONL as a third export format: Alongside CSV and XML, product exports now support JSONL - the format modern AI commerce protocols expect.
- Platform-specific feeds: Dedicated feeds can be created within the channel, starting with the OpenAI product feed connected to the OpenAI Merchant Center. More providers are expected to follow.
- Preconfigured feed template: All required fields come prefilled, invalid products are excluded automatically, special characters are handled cleanly.
- Tracking via affiliate infrastructure: Orders and visitors from AI platforms run through the existing affiliate/campaign fields - so you can see what the channel actually contributes economically.
Interesting for developers: the AbstractAgenticCommerceProductExportProvider provides a clean abstraction for implementing custom export providers for additional platforms. And important: the channel is flagged as experimental - expect adjustments in upcoming releases.
Why this is more than "yet another feed"
Product search no longer happens only on Google. More and more buying decisions start in ChatGPT, Copilot or Perplexity - with questions like "Which product fits my use case?". The answer comes from an AI that pulls product data from feeds and catalogs.
That changes the rules: in classic search you compete with SEO, ads and shop UX. In agentic commerce, your data quality competes. An AI does not click through your shopping experiences - it reads title, description, attributes, price and availability. Whatever is not in there cleanly does not exist for this channel.
The real task: your product data
The sales channel is set up in an hour. The honest work comes before that:
- Descriptions: Are they structured and factual - or marketing prose no AI can extract properties from?
- Variants: Are sizes, colors and versions modeled cleanly as variants, or hidden in free text?
- Availability and prices: Are stock levels and price logic correct in the export - including tier prices and promotions?
- Images: Are there meaningful main images in usable quality?
- Categories and attributes: Are properties maintained as structured data an agent can filter on?
If you have already done clean work for your Google Merchant Center feed, you have a clear head start: it is the same homework. Getting your product data in shape today serves Google Shopping, AI feeds and your own shop search at the same time.
What to do now - and what not to do
My assessment after the first days with the release:
- Do now: Review your product data quality and set up the feed as a test if you are on 6.7.10.0. The entry barrier is low, the learning effect is high.
- Watch: The channel is experimental. Do not build critical processes on it yet and expect changes to field and format details.
- Don't: Postpone the topic. The merchants whose data is clean today will be cited and recommended once AI channels gain volume - everyone else will retrofit under time pressure.
The rest of 6.7.10.0, by the way, is a classic stability release: roughly 117 resolved issues, mostly in checkout, search and product listings. One more reason not to fall too far behind on updates.