Skip to content
Sunnyvale Print
NVDA194.83 1.39%MSFT390.49 1.62%GOOGL359.91 0.36%AMZN242.67 0.40%META582.90 4.90%AAPL308.63 4.84%TSLA393.45 7.49%AMD517.82 4.26%AVGO360.45 2.41%PLTR129.30 2.84%ARM315.28 6.58%TSM434.16 2.27%NVDA194.83 1.39%MSFT390.49 1.62%GOOGL359.91 0.36%AMZN242.67 0.40%META582.90 4.90%AAPL308.63 4.84%TSLA393.45 7.49%AMD517.82 4.26%AVGO360.45 2.41%PLTR129.30 2.84%ARM315.28 6.58%TSM434.16 2.27%

Fetch MCP Server

The reference web-fetch server: give an agent a URL, get back the page as readable markdown. Small, honest, and easier to misuse than it looks.

MCP ServerVerifiedMedium riskopen-source

What it actually does

Another piece of the official reference collection, this one does a single job: fetch a web page and convert it to markdown an agent can actually read, with chunked retrieval for long pages. It's a Python package you run with uvx, and by default it respects robots.txt — a small courtesy most homegrown fetchers skip. The subtlety is where it runs: the request originates from your machine, so an agent can ask it to fetch things only your machine can see — localhost dashboards, internal services, cloud metadata endpoints. That's a feature right up until it's a finding.

Install

Read the source before you run this. Yes, actually.

From the desk

The Sunnyvale take

The kind of tool that looks too simple to review, which is precisely why you should. Forty lines of behavior, one real risk surface, and it does more day-to-day work than servers ten times its size.

Why it matters

Reading the live web is table stakes for a useful agent. This is the minimal, auditable version of that capability — the baseline before you reach for anything fancier.

In practice

An agent researching a library can pull the real docs page instead of hallucinating an API from training data — the difference between a guess and a citation.

The skeptic's note

It fetches one URL at a time with no search. If someone sells you a 'research agent' that is just this in a trench coat, you now know what you're paying for.

Use cases

  • Pull documentation into context on demand
  • Summarize articles and changelogs
  • Check what a page actually says before citing it

Risks & caveats

Requests originate from the host machine, so internal/localhost endpoints are reachable unless you sandbox it. Fetched page content is untrusted input to the model. Honors robots.txt by default; that default can be disabled, so check how yours is configured.

Related tools

Full index

Same shelf, different trade-offs.

MC-R01MCP Server

Filesystem MCP Server

Anthropic / MCP project · files
Verified

The reference server that gives an agent read/write access to your disk, fenced to directories you name. Simple, official, and the single sharpest tool in the drawer.

87.9k starsHigh riskChecked Jul 2, 2026
MC-R02MCP Server

Playwright MCP

Microsoft · browser-automation
Verified

Microsoft's browser-automation server — agents drive a real browser through the accessibility tree instead of squinting at screenshots. Fast, deterministic, and a genuinely good idea.

34.6k starsMedium riskChecked Jul 2, 2026