Cloudflare launches Web Search API for AI agents on AI Gateway

Cloudflare announced, on October 29, 2026, the beta launch of a Web Search API within AI Gateway. The news, reported by InfoQ, makes web search available via REST API and via Cloudflare Workers, with unified logging, access controls, and billing integrated into the same platform that already hosts thousands of AI applications.

For those building agents in Brazil, the detail that matters is the infrastructure: if the Worker already runs on Cloudflare, the agent gains native web search without contracting and integrating a separate search provider, without managing another loose API key in the project's .env, and without having to decide alone how to log those calls for auditing.

The problem the API tries to solve

The motivation behind the launch is a common behavior in AI agents today: when they need to fetch a web page, they often guess the URL and try a direct curl, hoping to get it right.

Michelle Chen (group product manager), Sam Else (senior director), and Gabriel Massadas (senior systems engineer), all from Cloudflare, describe the problem:

When you use an agent and it needs to fetch a live web page, the agent usually just guesses the URL of the page and then makes a tool call to curl it. This is why you'll sometimes see web fetches come back with a 404 Not Found, which happens if the agent incorrectly guesses the URL of that information. As you can imagine, it's not super efficient to randomly guess URLs all the time.

Michelle Chen, Sam Else and Gabriel Massadas, Cloudflare

It's exactly the scenario for those building RAG pipelines from scratch: the agent needs up-to-date context (a recent change in an API, a news story from the day, a piece of technical documentation) and, without a real search mechanism, is left at the mercy of trial and error.

How the API works

Applications and external services access web search through a REST endpoint, providing an AI Gateway authentication token and specifying the desired search provider in the request body. The endpoint returns results already formatted, ready to serve as context for a language model during inference.

Three providers are available at launch: Ceramic.ai, Exa, and Linkup. Cloudflare doesn't build its own search index here; it acts as a unification layer between these providers, billing usage together with the AI Gateway inference invoice and logging each search alongside model calls in the same logs.

The company also states that the crawlers used by search providers must follow Cloudflare's own publicly documented "verified bots" requirements, and that every search response must include a link to the source of the crawled content:

The crawler used by the web search provider must comply with Cloudflare's publicly stated requirements for "Verified bots" as defined in our developer documentation, and web search responses must include a link to the location of crawled content. These rules are net-positive for a fair Internet, and allow creators to decide what they want to do with their data.

Michelle Chen, Sam Else and Gabriel Massadas, Cloudflare

Pricing and data retention

Cloudflare says it will identify which providers offer Zero Data Retention and that search is billed at each provider's list price, with no additional markup from Cloudflare itself. For teams that already take compliance and LGPD (Brazil's data protection law) seriously in AI products, knowing in advance which providers don't retain search data is a direct selection criterion, without having to dig through terms of use.

The company also revealed that it plans to add native "server tools" to AI Gateway, starting with web search. Until then, anyone who wants to integrate search into their own workflow needs to do so via Cloudflare Workers.

What the technical community is questioning

A popular discussion on Hacker News debated search quality, data retention restrictions, and whether a unified API truly outperforms the alternatives already on the market. Simon Willison, creator of Datasette, summed up the question that matters most for those building products on top of it:

My number one question about search APIs is always if they allow you to store and resyndicate results you get from them. If I'm running an agent system but I'm not allowed to store the responses - or provide a "share transcript" button - that's a pretty significant limitation. The answer to that question is inevitably buried deep in the terms.

Simon Willison

It's a practical point: an agent that summarizes a search and shows the source to the end user can run into a contractual restriction from the provider behind the search, even when going through Cloudflare's layer. It's worth reading the terms of each provider (Ceramic.ai, Exa, Linkup) before designing a feature that depends on this.

Maximilian Goisser, CTO and co-founder of Erato Labs, sees the move from the perspective of vendor consolidation:

I think Cloudflare is (for companies already using it) approaching the status of trusted main cloud supplier (which usually would be AWS, GCP, Azure) via which the majority of cloud costs are billed (so you don't have to go through a fresh procurement process).

Maximilian Goisser, CTO and co-founder of Erato Labs

What changes for developers in Brazil

For Brazilian teams that already use Cloudflare Workers as an edge layer, the Web Search API reduces a real dependency: today, building RAG with live search usually means subscribing to OpenAI's, Anthropic's, or a third-party provider's search API, and dealing with billing, authentication, and logging in systems separate from the rest of the stack.

With Cloudflare's API inside AI Gateway, that process falls under the same bill, the same observability dashboard, and the same authentication flow that already exists for model inference. This makes it easier, for example, to audit how much an AI application spent on web search versus model calls in a single place, without reconciling two invoices.

The thing to watch is that Cloudflare doesn't bring its own search index: it depends on Ceramic.ai, Exa, and Linkup, providers less known in the Brazilian market than Google or Bing. Before migrating a production pipeline to the new API, it's worth comparing the quality and coverage of these providers' results for the application's specific language and domain, and confirming which one offers Zero Data Retention if that's a compliance requirement for the project.