Cloudflare Extends AI Search to Make it Easier for Agents and Developers to Search Custom Data

August 30, 2026 | Source: InfoQ AI/ML

Tags: Cloudflare, AI Search, RAG, vector database, MCP, AI agents, developer tools

Cloudflare AI Search packages its existing vector DB, crawler, and embedding primitives into a single CLI command, giving AI agents a managed search engine over custom data — with free default embeddings, automatic page discovery, and public MCP endpoint support.

Details

Cloudflare has shipped AI Search, a managed search-and-retrieval layer that chains together Workers AI, Vectorize, R2, AI Gateway, and Browser Run into a single deployable unit. Rather than wiring these primitives together manually, a single `npx wrangler ai-search create` command handles crawling, ingestion, embedding, and retrieval end-to-end. The service ships with two crawl modes: standard mode indexes sites via their sitemap, while the new \'discover\' mode traverses pages autonomously even without one. A single public endpoint can span multiple data sources simultaneously — Cloudflare's own API docs, Astro, Vite, Hono, and Replicate documentation are already indexed as a unified corpus. For agent integration, AI Search exposes public `/mcp` and `/search` endpoints requiring no authentication or deployment, making it drop-in compatible with existing agent toolchains and MCP servers. Cloudflare also supports Worker-based integration for teams that want more control. Cloudflare states that embedding and re-ranking are free when using the default model, with pricing designed to be predictable and scalable — though specific tiers for custom models or high-volume usage are not detailed in the source.