llms.txt can be useful as a concise map for documentation and stable resources. It is not a replacement for a live commerce surface. Prices change, stock changes, variants differ and delivery constraints are contextual. An agent that wants to buy needs a current answer from the store.Two artifacts, two jobs
An llms.txt file is a static orientation document. It can point an AI system toward useful pages, describe terminology and reduce the work of navigating a documentation site. That is a sensible use of a small, human-maintained summary.
A product catalog has another job. It must answer questions that are only true at request time: what is the current price, which variation is in stock, what is the purchase URL, and what facts apply to this exact offer? Copying those answers into a static text file turns them stale as soon as the store changes.
Orientation for stable documentation.
- Links and context
- Editorially maintained
- Useful for a site map
- Can be static
Verification for an active commercial offer.
- Search and product lookup
- Current price and availability
- Variants and canonical purchase URL
- Must read the store's present state
The freshness problem is not theoretical
Even a well-generated export is a snapshot. It may be useful for a feed whose delivery rules are explicit, but it is a poor authority for a decision made after a price update or an inventory change. The answer is not to reject static documents; it is to keep them in the layer where they belong.
For WooCommerce, a stronger pattern is: publish stable documentation where it helps, announce a structured catalog route, then let the agent verify live operational facts before sending the customer to checkout. That keeps the data close to the merchant who owns it.
What Bridge does — and does not — do
KaliCart Bridge does not generate an llms.txt file for a merchant catalog. It exposes a bounded, read-only live catalog surface on the WooCommerce domain, with discovery signals and product search/lookup for compatible agents. This is intentionally different from a static snapshot.
The distinction also keeps claims honest. A live endpoint does not force an LLM to use it; it gives a compatible client a better way to verify the facts when it does. See WooCommerce Product API for AI Agents for the implementation and why compact product data matters.