Free web page reader MCP servers, tested

We sent the same 3 questions with known answers to 6 free web page reader MCP servers on 2026-11-14, as read-only calls without an API key. All of them pass our daily live check. 2 gave answers you could rely on. Winner: Exa (2/3, 2.3 s).

The questions

en.wikipedia.org/wiki/Model_Context_Protocol → correct answer: introduced by Anthropic in November 2024 (checked at Wikipedia)
kitscout.dev/t/view.html?... (JavaScript-only) → correct answer: the tool's description (rendered by JS) (checked at our test page)
arxiv.org/pdf/1706.03762 → correct answer: Attention Is All You Need (checked at arXiv PDF)

Results

#	Server	Q1	Q2	Q3	Score	Avg
1	Exa `https://mcp.exa.ai/mcp`	PASS	FAIL	PASS	2/3	2.3 s
2	Clearcrawl `https://agentreader.onrender.com/clearcrawl/mcp`	PASS	FAIL	PASS	2/3	5.5 s
3	Vanshul Markdown `https://mcp.vanshul.com/mcp`	PASS	FAIL	FAIL	1/3	0.3 s
4	search-mcp (ioconnex) `https://search.ioconnex.dev/mcp`	PASS	FAIL	FAIL	1/3	0.6 s
5	Page Fetch `https://pagefetch.openkrill.app/mcp`	PASS	FAIL	FAIL	1/3	0.7 s
6	TidyTools `https://tools.yukai.uk/mcp`	PASS	FAIL	FAIL	1/3	0.8 s

Same questions with known answers sent to every server on the same day as read-only tool calls (no API key). PASS = the correct answer appears in the top results; HALF = partly right or outdated source; FAIL = wrong; NO REPLY = request refused. Results can change; live status is re-checked daily, tests are re-run periodically.

Evidence: what each server returned

Excerpts of the servers' own replies, quoted as returned.

1. Exa · Read the PDF. Admitted the JS page was empty

PASS web_fetch_exa("en.wikipedia.org/wiki/Model_Context_Protocol")
"...introduced by Anthropic in November 2024..."
FAIL web_fetch_exa("kitscout.dev/t/view.html?... (JavaScript-only)")
Error: CRAWL_EMPTY_CONTENT (no content, and it said so)
PASS web_fetch_exa("arxiv.org/pdf/1706.03762")
"# Attention Is All You Need"

2. Clearcrawl · Reads PDFs, but slow

PASS read_url("en.wikipedia.org/wiki/Model_Context_Protocol")
"...introduced by Anthropic in November 2024..."
FAIL read_url("kitscout.dev/t/view.html?... (JavaScript-only)")
"markdown": "[KitScout] [Categories] [Platforms]..." (menu only)
PASS read_url("arxiv.org/pdf/1706.03762")
"kind": "pdf", "pages": 15 ... "Attention Is All You Need"

3. Vanshul Markdown · Fastest, HTML only

PASS fetch_markdown("en.wikipedia.org/wiki/Model_Context_Protocol")
"...introduced by [Anthropic] in November 2024..."
FAIL fetch_markdown("kitscout.dev/t/view.html?... (JavaScript-only)")
"... [Field tests] [For agents] Loading..."
FAIL fetch_markdown("arxiv.org/pdf/1706.03762")
Error: Page is not a text or HTML document (got application/pdf)

4. search-mcp (ioconnex) · Plain text, HTML only

PASS fetch_url("en.wikipedia.org/wiki/Model_Context_Protocol")
"...introduced by Anthropic in November 2024..."
FAIL fetch_url("kitscout.dev/t/view.html?... (JavaScript-only)")
"... Field tests For agents Loading..."
FAIL fetch_url("arxiv.org/pdf/1706.03762")
unsupported content type 'application/pdf'

5. Page Fetch · Clean Markdown, HTML only

PASS fetch_page("en.wikipedia.org/wiki/Model_Context_Protocol")
"...introduced by [Anthropic] in November 2024..."
FAIL fetch_page("kitscout.dev/t/view.html?... (JavaScript-only)")
title: Tool — KitScout description: Agent tool details. (no page content)
FAIL fetch_page("arxiv.org/pdf/1706.03762")
status: skipped reason: not_html

6. TidyTools · Free quota ran out

PASS read_url("en.wikipedia.org/wiki/Model_Context_Protocol")
"...introduced by Anthropic in November 2024..."
FAIL read_url("kitscout.dev/t/view.html?... (JavaScript-only)")
"Paid browser rendering: https://tools.yukai.uk/..."
FAIL read_url("arxiv.org/pdf/1706.03762")
"This source's daily free allowance is exhausted."