Web fetch

Overview

Tool identifier web_fetch. web_fetch does a plain HTTP GET on a given URL and extracts its readable content: HTML is converted to markdown or plain text, and PDFs are extracted to text via pypdf. It does not execute JavaScript, so pages that render entirely client-side or require login may come back incomplete. Unlike link_reader, this tool needs no credentials of its own (it is a plain HTTP fetch) — use it to let the agent read articles, docs, or any public URL the user references. Import path: from veadk.tools.builtin_tools.web_fetch import web_fetch

Environment & prerequisites

Environment variables:

MODEL_AGENT_API_KEY: API key for the agent’s reasoning model (the web_fetch tool itself needs no extra credentials)

Usage

Parameters:

url: the http(s) URL to fetch;
extract_mode: markdown (default, keeps headings / links / lists) or text (plain text);
max_chars: maximum characters of extracted content (default 50000).

Returns {"url", "title", "content", "truncated"}, or {"error": ...} on failure.

examples/tools/web_fetch/agent.py

import asyncio

from veadk import Agent, Runner
from veadk.memory.short_term_memory import ShortTermMemory
from veadk.tools.builtin_tools.web_fetch import web_fetch

agent = Agent(
    name="web_fetch_agent",
    model_name="doubao-seed-2-1-pro-260628",
    description="An agent that reads web pages and PDFs.",
    instruction="Use the web_fetch tool to fetch the given URL, then answer based on its content.",
    tools=[web_fetch],
)

runner = Runner(agent=agent, short_term_memory=ShortTermMemory())


async def main():
    response = await runner.run(
        "Fetch https://arxiv.org/pdf/1706.03762 and summarize the paper's core idea"
    )
    print(response)


if __name__ == "__main__":
    asyncio.run(main())

Notes

Security & limits:

SSRF protection: after DNS resolution it blocks private / loopback / link-local / reserved addresses, and re-validates every redirect hop (including <meta refresh>), following at most 3 hops.
Limits: 2 MB download cap for HTML (10 MB for PDFs); 30 s request timeout; results cached in-process for 15 minutes.
No JavaScript rendering; hostname checks do not replace network access restrictions in the deployment environment.

Direct calls and parameter limits

Parameter	Type	Default	Description
`url`	`str`	Required	Public HTTP or HTTPS URL
`extract_mode`	`str`	`"markdown"`	`markdown` or `text`; other values use `markdown`
`max_chars`	`int`	`50000`	Output character limit, clamped to 1–200000; unconvertible values use the default
`tool_context`	`ToolContext \| None`	`None`	Injected for agent calls; optional for direct calls

Direct calls need no model API key. PDF extraction requires pypdf and does not perform OCR on scanned pages. truncated only indicates that extracted text exceeded the character limit; it does not guarantee a complete download or full content coverage.

from veadk.tools.builtin_tools.web_fetch import web_fetch

result = web_fetch("https://example.com", extract_mode="text", max_chars=2000)
if "error" in result:
    print("Fetch failed:", result["error"])
else:
    print(result["url"], result["title"])
    print(result["content"])
    print("Truncated:", result["truncated"])

Overview

Get started

Core components

Observability

Extensions

Frontend

Security

Deployment

Overview

Environment & prerequisites

Usage

Notes

Direct calls and parameter limits

​Overview

​Environment & prerequisites

​Usage

​Notes

​Direct calls and parameter limits

Overview

Environment & prerequisites

Usage

Notes

Direct calls and parameter limits