Web Fetch

The Web Fetch extension gives the agent the ability to fetch and read public web pages. A single clean tool: web_fetch.

What It Does

The agent can fetch any public URL and return its content in the most useful format:

Format	Use Case
markdown (default)	Read documentation, blog posts, articles
text	Extract plain text content
html	Get raw HTML for analysis
json	Fetch JSON APIs directly

Usage

web_fetch {
  url: "https://docs.example.com/api-reference",
  format: "markdown"
}

The tool handles redirects automatically and rejects binary content (images, PDFs, etc.). It enforces SSRF protection — private and internal IPs are blocked.

Typical Workflows

Reading Documentation

The agent can fetch documentation pages to understand APIs, libraries, and frameworks before writing code:

"Check the Next.js 14 routing docs and update our layout based on the latest patterns."

Consuming API References

Fetch JSON endpoints directly:

web_fetch {
  url: "https://api.github.com/repos/owner/repo/releases/latest",
  format: "json"
}

Research & Context

Before implementing a feature, the agent can fetch relevant blog posts, Stack Overflow answers, or RFC documents to gather context.

Safety

SSRF protection — private/internal IPs (127.0.0.1, 10.x, 192.168.x, etc.) are blocked
HTTP/HTTPS only — no file:// or other protocols
Timeout — configurable up to 120 seconds (default 30s)
Binary rejection — non-text content is rejected automatically

Configuration

Setting	Default	Description
`timeout`	`30`	Request timeout in seconds (max 120)
`userAgent`	Default	User agent string for requests

When to Use

Reading documentation — fetch library docs before writing integrations
API exploration — inspect REST/JSON endpoints
Research — gather context from web articles
Content extraction — read blog posts, READMEs, changelogs