WebFetch

A built-in tool that fetches a URL and returns its text content. Allows agents to browse the web.

Usage

import { agent, WebFetch } from '@daedalus-ai-dev/ai-sdk';

const response = await agent({
  instructions: 'You summarize web pages concisely.',
  tools: [new WebFetch()],
}).prompt('What is https://example.com about?');

console.log(response.text);

Behaviour

• Sends a GET request with User-Agent: daedalus-ai-sdk/0.1
• Returns the raw response body as text (HTML, JSON, plain text, etc.)
• Truncates to 10,000 characters with a … [truncated] suffix to avoid overwhelming the context window
• Throws on network errors (the SDK passes the error to the model)

Tool definition

The tool registers itself as:

Field	Value
Name	web_fetch
Description	Fetch the content of a URL and return it as text. Use this to retrieve web pages, APIs, or any HTTP resource.
Parameter	url (required string)

Limitations

• No JavaScript rendering. WebFetch fetches raw HTML — it does not execute JavaScript. Pages that require JS to display content will return empty or minimal HTML.
• No authentication. For authenticated endpoints, build a custom tool that adds the necessary headers or credentials.
• Rate limits. Calling WebFetch many times in parallel may trigger rate limiting on the target server.

Custom web tool

For more control, define your own web fetcher:

import { defineTool } from '@daedalus-ai-dev/ai-sdk';

const authenticatedFetch = defineTool({
  name: 'fetch_api',
  description: 'Fetch data from our internal API.',
  schema: (s) => ({
    endpoint: s.string().description('API endpoint path, e.g. /users/123').required(),
  }),
  handle: async (input) => {
    const res = await fetch(`https://api.internal.com${input.endpoint}`, {
      headers: {
        Authorization: `Bearer ${process.env.INTERNAL_API_KEY}`,
        Accept: 'application/json',
      },
    });

    if (!res.ok) throw new Error(`API error: ${res.status}`);
    return await res.text();
  },
});