← Back to Knowledge Base
Affect Performance Team
|News|Nov 17, 2026

Storyblok Introduces Strata Semantic Search for Content and AI Tools

On October 26, 2026, Storyblok introduced Strata, a semantic search system for content stored in its headless CMS. Instead of relying only on exact keyword matches, Strata finds stories based on the meaning of a query. It can be accessed through the Content Delivery API, Management API, and Storyblok MCP Server, allowing developers to use content search on websites, in applications, and through compatible AI assistants.

Strata indexes Storyblok content as vector embeddings, numerical representations that capture semantic relationships. It compares the meaning of a query with the indexed content and ranks the closest matches, even when a visitor uses different words from those in a page.

Semantic Search Beyond Exact Keyword Matching

Strata adds meaning-based retrieval to the content discovery options available to Storyblok customers. Traditional keyword search finds documents containing specific words. Fuzzy search accounts for spelling mistakes and predictable wording variations. Semantic search can connect different phrases that express the same underlying need.

For example, a visitor searching for “how do I send an order back?” may find a page called “Return Policy,” despite the absence of an exact phrase match. An ecommerce search for “clothes for cold weather” might surface articles about winter jackets, insulated accessories, or other relevant items.

Strata compares the query embedding with indexed stories and ranks results by semantic similarity. It does not, by itself, generate an answer. It returns relevant stories and content excerpts that a website, application, or AI assistant can process further.

Search Through the Content Delivery API

The main integration point for public-facing applications is the Strata search endpoint in the Storyblok Content Delivery API:

GET https://api.storyblok.com/v2/cdn/strata/search

The q parameter accepts a natural-language search string. Developers can also specify the content version, result limit, content type, and story URL prefix. For example, content_type=page limits retrieval to stories of the page content type, while starts_with narrows the search to a specific part of a site.

The API returns Storyblok stories ordered by relevance. Each result includes story information such as its ID, name, and slug, along with a matching excerpt and a score indicating relevance.

The default result limit is 25. The maximum is 100, reduced to 50 when a content_type or starts_with filter is applied. The same story can appear more than once when multiple passages match the search.

This response structure lets developers control result grouping, the search interface, and what happens when a visitor selects an item.

Separate Search for Draft and Published Stories

Strata maintains separate search indexes for draft and published content. This distinction allows public sites to search only released pages while internal editorial tools can retrieve work in progress.

In the Content Delivery API, the version parameter and access token determine which content can be retrieved. A public access token is used for published stories, while draft retrieval requires an appropriate preview token.

Access restrictions are respected. A token scoped to particular content cannot search beyond that scope. This matters for organizations with multiple editorial approval stages: the presence of a draft in the CMS must not automatically expose it to public website visitors or external AI applications.

Private Content Search With the Management API

Strata is also available through Storyblok’s Management API, which is intended for internal tools that act on behalf of authenticated users:

GET https://mapi.storyblok.com/v1/spaces/:space_id/strata/search

The Management API accepts personal access tokens or OAuth authentication. Search results are restricted to stories the authenticated user can read; a scoped token also needs the strata:read permission.

This route is suited to editorial applications and other private workflows. An editor can search for existing content before writing a new page, while an internal application can find related resources or reusable content components.

Storyblok notes that this endpoint is more tightly rate-limited than the Content Delivery API. For high-volume public search, it recommends using the Content Delivery API instead.

Multilingual Queries and Default-Language Indexing

Strata accepts search queries in languages other than the language of the indexed content. For example, a user could search in German or Russian and receive matches from English-language stories when the meanings align.

There is an important limitation: the current Strata index covers only the default language of a Storyblok space. Translated story versions do not form separate indexes. Multilingual querying therefore does not mean the system searches across all localized versions of a website.

If an application needs to display matching content in the visitor’s preferred language, the developer must separately retrieve and render the appropriate localized version.

Automatic Indexing and Manual Reindexing

After a space administrator enables Strata under Settings → Strata, the platform performs an initial full indexing run. Storyblok estimates approximately one to two minutes per 1,000 stories, depending on content size and structure. Search becomes available after that first run finishes.

Strata refreshes each space’s index automatically once per day. Administrators can start a manual reindex from the Index Status panel and choose all stories, only drafts, or only published stories. Subsequent runs generally process stories changed since the previous indexing job and are usually faster.

Search remains available during reindexing using the existing index, but newly published or modified stories might not appear until the refresh completes. The administration interface also displays a run history and errors for stories that could not be indexed.

Storyblok MCP Server and AI Assistants

Strata works with Storyblok MCP Server, which exposes supported CMS capabilities to compatible AI assistants through Model Context Protocol. An assistant can use natural-language requests to search a Strata-enabled Storyblok space.

For example, a content strategist can ask an assistant to locate articles on a product category, find pages related to a campaign topic, or identify existing material before planning new content. The MCP server passes the request to Storyblok and returns search results for use in the conversation.

The distinction between retrieval and action is important. Strata finds relevant content; it does not independently edit, classify, publish, or fact-check the stories. Any subsequent CMS operations must be implemented through other authorized tools and, where appropriate, confirmed by the user.

Contextual Recommendations and Retrieval-Augmented Generation

Storyblok also presents Strata as a foundation for contextual recommendations and retrieval-augmented generation (RAG). Instead of passing only a short search phrase, an application can submit context about a product, audience, or currently viewed page to find semantically related stories.

An ecommerce product page could use category and product details to retrieve relevant guides, comparisons, or usage recommendations. A B2B website could similarly surface case studies, product documentation, or articles related to the page a visitor is reading.

In a RAG workflow, Strata retrieves relevant CMS stories and excerpts. A separate language model can then use those materials as context when composing a response, reducing its dependence on general or outdated model knowledge.

Strata is not a standalone personalization engine or answer generator. The application still needs to implement recommendation logic, context management, answer generation, and validation.

Availability and Indexing Limitations

At launch, Strata is available on Storyblok’s Elite plan for one space per organization. A space administrator must enable it through Settings → Strata; it is disabled by default. The settings interface also includes a test search after indexing completes.

The index currently covers stories on the default content branch and in the space’s default language. The asset library, separate datasources, and tags are not indexed as independent searchable objects. Text contained in story fields, including image alt text, can be indexed.

The October 26 announcement describes the Elite-plan availability but does not explicitly label the release as a public beta or as generally available. Strata also does not automatically assign tags, classify content, fact-check material, or generate new text.

About Storyblok Strata and Headless Content Management

Storyblok is a headless CMS for managing structured content across websites, applications, and other digital channels. It combines visual editing, component-based architecture, and APIs for content delivery.

Strata adds semantic retrieval to that architecture. Ecommerce teams can use it to build search across help articles, product content, and editorial catalogs. Service companies can use it for knowledge bases, and B2B teams can retrieve case studies, technical documentation, and product resources.

Because Strata is available through the Content Delivery API, Management API, and MCP Server, the same indexed content can support public discovery, private editorial searches, and AI-driven applications. Generating answers, personalizing experiences, or changing content still requires additional tools and application logic.

Trust Us Transform Your Performance Marketing.

Connect better content discovery with measurable acquisition. We audit your website analytics, identify where users struggle to find information, and turn those insights into a practical conversion optimization plan.

Audit Your Website Analytics

Sources