Cloudflare AI Search is a built-in search and retrieval service designed to give AI agents and applications a ready-to-use search engine over custom data. It supports agent integration, multimodal search, and seamless integration with other Cloudflare tools.
A search pipeline typically involves several components, including a crawler, parser, embedding model, vector database, search API, and more. Cloudflare already provides many of these as "primitives", including Workers AI, AI Gateway, Vectorize, R2, Browser Run, and others. AI Search now brings these capabilities together, automatically handling the end-to-end search pipeline and doing so "better", the company says.
Our goal is to give your agents their own search engine, where they can easily find data to provide better answers for themselves and their humans.
AI Search enables organizations to index collections of structured or unstructured data ans make them accessible to agents. Previously, each indexed website was required to publish a sitemap, but AI Search is now able to work in "discover" mode, automatically discovering pages even without one. The engine provides a single public endpoint for searching across multiple instances or websites at once without requiring authentication. For websites built using the open-source CMS EmDash, AI Search can be easlily integrated using a specific plugin.
To create a search instance, you run a single command that handles crawling, ingestion, embedding, and retrieval:
npx wrangler ai-search create cloudflare-community \
--namespace dev-stack \
--source https://community.cloudflare.com \
--type web-crawler \
--parse-type discover
Cloudflare AI Search is available across several Clouflare-managed services, including its API and developer documentation, Astro, Vite, Hono, Replicate, and others. This allows these sources to be indexed and searched as a single corpus rather than as independent entities, enabling a single query to answer a question using data from the different sources. Search can be integrated into an existing application or MCP server through a Worker, which Cloudflare recommends, or exposed through public /mcp
and /search
endpoints that require no authentication or deployment and can be easily shared with others.
As a final note about AI Search, Cloudflare emphasizes that it designed its pricing model to make it predictable and scalable. In particular, embedding and re-ranking are free when using the default models or select models from the Workers AI catalog, while answer generation and query rewriting are billed based on model usage. This pricing model will take effect once the service reaches general availability, while during the beta period, AI Search is available for free.