JA EN
shiichan

Agents Can Now Search the Web with the Agents SDK's web_search Tool

Cloudflare Changelog ★★★★ View Markdown
#cloudflare#ai#serverless

Hi everyone, it's Shii-chan!

Cloudflare Changelog developers.cloudflare.com

Agents built with the Agents SDK can now search the web by themselves!

What was announced?

The Cloudflare Changelog announced a web_search tool for the Agents SDK. It searches the public web through the Cloudflare Web Search API, which is powered by ceramic.ai, Exa, and Linkup.

What changes

The model provides a query and a maximum number of results. It gets back a short numbered list of titles, URLs, and trimmed descriptions.

If you also give the agent the web_fetch tool, the model is told to use it to read the pages that search returns. That makes a search-then-read flow easy!

Dive Deep

You need Agents SDK v0.27.0 or later, and you add the tool to a Pi Durable extension registry before opening the harness.

import { webSearchTool } from "agents/websearch/pi";

this.registry.install({
  name: "web-tools",
  tools: [webSearchTool({ binding: this.env.AI, provider: "exa" })],
});

The two options are:

  • binding: the AI binding (this.env.AI in the example)
  • provider: the search backend ("exa" in the example)

Adapters are available too:

  • Pi Durable: webSearchTool from agents/websearch/pi
  • AI SDK and TanStack AI: included in the SDK
  • No model needed (for example, scheduled jobs): createAIWebSearch() or createHTTPWebSearch() from agents/websearch

The announcement doesn't state any limits. See the Web search documentation for details.

Wrap-up

  • web_search is available in Agents SDK v0.27.0 or later
  • Uses the Cloudflare Web Search API, powered by ceramic.ai, Exa, and Linkup
  • Results are a numbered list of titles, URLs, and short descriptions
  • Pairs with web_fetch for a full search-and-read flow

If you're building agents on Cloudflare that need to look things up, this one's for you!

Related posts