Storyblok Strata adds semantic search for CMS stories, with Elite and indexing limits

Storyblok added Strata in an [October 26 changelog update](https://www.storyblok.com/cl/2026‑october‑strata), giving developers a way to retrieve CMS stories by meaning rather than exact keywords. The feature is available on the Elite plan for one space per organization. It can supply content for site search, chatbots and retrieval‑augmented generation applications. The launch exposes semantic search through both the Content Delivery API and the Management API. It also lets MCP clients, including Claude, search a space where Strata is enabled through the Storyblok MCP server. For a team building an assistant around its existing content, the useful change is a retrieval route into that CMS collection. Strata supplies matching stories; the application still determines what to do with them.

The [Strata concept documentation](https://www.storyblok.com/docs/concepts/strata) describes an index of story content represented as vector embeddings. A query is compared with that content to retrieve relevant matches. Strata does not itself write an answer or fact‑check the material it finds. An application can pass retrieved stories to an AI model as context, but that is a further step rather than a built‑in guarantee about the resulting answer. The same distinction applies to personalization. Developers can submit audience or product context as a query to find related stories, as the [launch note](https://www.storyblok.com/cl/2026‑october‑strata) describes. That makes Strata a source of inputs for a recommendation workflow; it does not automatically personalize an experience in real time. The documentation also excludes content generation, rewriting and automatic tagging from Strata’s role.

The [setup manual](https://www.storyblok.com/docs/manuals/strata) narrows the searchable collection to stories on the default branch and in the default language. Draft and published stories have separate indexes. The asset library, datasources and tags are excluded, although text held in an asset field inside a story, such as alt text, can be indexed. Queries in another language still search the default‑language content, rather than adding every translation to the index. Access also depends on the token. A public or preview Content Delivery API token selects the corresponding content version. Personal access and OAuth tokens through the Management API search within the user’s permissions. The [concept documentation](https://www.storyblok.com/docs/concepts/strata) distinguishes the high‑volume Content Delivery API route from lower‑volume, rate‑limited private Management API queries. A semantic query does not remove those access boundaries.

Strata is disabled by default. A space administrator must enable it under Settings → Strata, and the initial indexing run must finish before searches return results, according to the [manual](https://www.storyblok.com/docs/manuals/strata). The service refreshes its index on a daily schedule and also supports manual reindexing. The current index continues serving searches during a reindex, so publication and search visibility need not happen at the same moment. For an Elite customer evaluating the release, the deployment questions are concrete: which space to enable, which content version and permissions an application should use, and how its freshness requirements fit the indexing schedule. The official material establishes those supported routes and limits. It does not provide an independent test of search quality or downstream AI‑answer accuracy. *AI‑generated conceptual illustration by OpenTools Team: separate document trays, text‑like story cards and a magnifying lens represent retrieval from indexed content collections. This is illustrative artwork, not an actual Storyblok Strata interface, architecture diagram or search result.*