Log inSubmit server
Lookout logo

Lookout

Searches, fetches, extracts, caches, crawls, researches, and compares web content with source spans and warnings.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install Lookout

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "lookout": {
      "command": "npx",
      "args": [
        "-y",
        "@sylphx/anymd"
      ]
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

We couldn’t confirm it starts correctly

We ran the install command but it didn't respond within our test window — this can mean a slow first-time install rather than a real problem.

npx -y @sylphx/anymd

No response to initialize.

This is an experimental automated check and can have false negatives — missing environment variables, a slow cold install, etc. It doesn’t necessarily mean something’s wrong. Last checked 19d ago.

Overview

Lookout MCP server gives AI agents tools for web search, URL fetching, HTML extraction, caching, bounded crawling, multi-page research, and text comparison. Search uses DuckDuckGo HTML, Wikipedia OpenSearch, npm, and Hacker News Algolia, while fetch and extract return source-level spans, metadata, tables, warnings, and gaps instead of generated summaries. Use it when an agent needs inspectable web evidence without an API key; install it with npx as a stdio server. The repository was archived after being merged into anymd on September 25, 2026.

Use cases

  • Search public sources and inspect the underlying pages
  • Collect cited excerpts from several top-ranked results
  • Compare changes between two web pages or texts
  • Crawl a bounded same-origin site without browser rendering

Key features

  • Search across four public web adapters
  • Fetch URLs with SSRF and response limits
  • Extract metadata, tables, and source spans
  • Cache search and fetch responses on disk
  • Run bounded same-origin crawls
  • Compare added and removed words

Tools7

What Lookout lets your agent do. Expand a tool for its parameters.

Self-reported~121 tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

Not checked yetLast checked 2d ago

We have not completed a health check for this listing yet.

0/9 checks healthy over the last 44d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 14d ago via OSV.dev · @sylphx/anymd (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars1Stargazers on the source repository.
npm downloads20kPackage downloads in the last 30 days.
Last commit1mo agoMost recent push to the default branch.
Install checkInconclusiveDidn't respond in our test window — often a slow first install.
Tools exposed7Callable tools this server registers over MCP.
Directory activity2 viewsConfig copies, upvotes, and views on AllMCPs.

How Lookout works

What Lookout MCP server does

The Lookout MCP server exposes web research operations through MCP for agents that need more than search-result snippets. Its tools can find ranked results, retrieve one URL, extract readable page details, query a disk cache, crawl within one origin, research several top results, and compare two texts or URLs.

Search combines four public sources: DuckDuckGo HTML, Wikipedia OpenSearch, the npm registry, and Hacker News Algolia. Results identify their adapter and are ranked, but search does not automatically open the returned pages.

Returned evidence can include titles, descriptions, excerpts, metadata, tables, and spans. Failures remain visible through warnings or errors for blocked addresses, HTTP failures, truncated responses, or empty extraction results. The server does not turn missing content into a polished model-generated answer.

How it works

The Lookout MCP server runs as a stdio process. A client starts the package and communicates with the exposed MCP tools. The web_research tool provides a named sequence that searches, then fetches and extracts the highest-ranking pages. It uses three pages by default and can process up to six.

Fetching reads HTTP responses rather than rendering pages in a browser, so JavaScript-dependent content is not executed. Requests apply SSRF protections, a redirect limit, a response-size limit, and a body-prefix span. Private, loopback, and link-local addresses are rejected.

Search and fetch use the on-disk cache by default. A cache hit replays an earlier response, while the first search still requires network access. Passing useCache: false disables that behavior for the relevant operation. The cache tool manages or queries this stored data and is not itself a search tool.

Setup and configuration

Install and start the Lookout MCP server with:

Terminal
npx -y @sylphx/lookout

No API key is required. For Claude Code, the documented command is:

Terminal
claude mcp add lookout -- npx -y @sylphx/lookout

Claude Desktop, Cursor, VS Code, and Codex can use a stdio configuration with npx as the command and -y, @sylphx/lookout as its arguments. The repository also documents development commands using Bun, including tests and a diagnostic command.

Tools and capabilities

  • web_search: ranked results from four public adapters.
  • web_fetch: retrieves one URL with address checks, redirect and size limits, and a source span.
  • web_extract: reads titles, metadata, tables, and spans from HTML or a URL.
  • web_cache: queries or manages the disk cache.
  • web_crawl: performs a bounded crawl restricted to the same origin.
  • web_research: searches, fetches, and extracts multiple top pages.
  • web_diff: reports added and removed words between texts or URLs.

Limitations and notes

The Lookout MCP server does not render a browser or synthesize an answer. web_fetch does not honor robots.txt unless respectRobots: true is set; web_crawl honors it by default unless that behavior is disabled. A failed adapter, non-200 response, truncated body, or empty extraction is reported rather than concealed.

This repository is archived and was merged into anymd on September 25, 2026. The README identifies the package and server under the MIT license. Its documented scope is web content; companion products listed in the README handle PDFs, images, video, repositories, and code retrieval separately.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Lookout

Lookout is an MCP server that connects AI agents to public web search sources and individual URLs. Its main tools search, fetch, extract, cache, crawl, research, and compare web content. Results can include source spans, metadata, tables, warnings, and gaps, without requiring an API key or generating a synthetic answer.

Install it with `npx -y @sylphx/lookout`, which starts a stdio MCP server. No API key is required. Claude Code can use `claude mcp add lookout -- npx -y @sylphx/lookout`; other documented clients use an equivalent stdio configuration.

Yes, Cursor is listed as a supported setup target. Configure it as a stdio MCP server with `npx` as the command and `-y`, `@sylphx/lookout` as the arguments.

No, Lookout fetches HTTP responses and does not run JavaScript or render a browser. Pages whose content depends on client-side rendering may therefore not provide the expected extracted material.

No, the documented search adapters are public and the installation requires no API key. Search uses DuckDuckGo HTML, Wikipedia OpenSearch, the npm registry, and Hacker News Algolia.

Related MCP servers

All in Search & Data Extraction All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewLookout AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/lookout?style=directory)](https://allmcps.com/mcp/lookout)
HTML Embed
<a href="https://allmcps.com/mcp/lookout"><img src="https://allmcps.com/api/badge/lookout?style=directory" alt="Lookout on AllMCPs" /></a>