Research

Apify's MCP server exposing thousands of ready-made scrapers for social media, search engines, maps, and e-commerce.

Why write a scraper when five thousand already exist? Rent the long tail instead.

Local MCP server for arXiv literature work: search, section-level reads of original LaTeX, BibTeX, topic watches.

Papers stay on disk and Claude reads the LaTeX, not a mangled PDF. Serious literature-review kit.

Brave's official MCP server for the Brave Search API: web, news, image, and local search.

Replaced the archived reference server, and it's the cheapest way to give Claude a real search index.

Web search through DuckDuckGo with content fetching and parsing. No API key required.

The zero-signup option. When you refuse to manage another API key, this is the fallback that works.

Exa's official MCP server: semantic web search and crawling built for AI agents, with company and code search.

Search built for models, not humans. The results read like context, not ten blue links.

Web scraping and search MCP server from Firecrawl. Scrape, crawl, and extract structured data from any site.

The default answer when Claude needs a whole site, not a page. Handles the JavaScript mess for you.

Kagi's official MCP server for Kagi Search and summarizer tools.

Ad-free search results are quietly better inputs for a model. Kagi subscribers should not skip this.

Turns Claude into a data-scientist assistant: load CSVs and run exploratory analysis scripts against them.

Point it at a mystery CSV and let Claude do the tedious first hour of every analysis.

Reference MCP server that fetches web pages and converts them to markdown for efficient LLM consumption.

The plainest tool in the official repo and still one of the most used. URL in, markdown out.

Retrieves transcripts from YouTube videos for summarization and analysis.

Half the internet's knowledge is trapped in hour-long videos. This is the crowbar.

Searches and downloads papers across arXiv, PubMed, bioRxiv, Semantic Scholar, OpenAlex, Crossref, and more.

Twenty-odd academic sources behind one tool call. The shotgun approach to literature search, usefully so.

Official Perplexity MCP server adding real-time web search, reasoning, and deep research via the Sonar API.

An answer engine inside your answer engine. Oddly effective for questions Claude would otherwise hedge on.

Search, access, and analyze PubMed biomedical articles through MCP, including metadata and PDF retrieval.

Narrow and deep beats broad and shallow for biomedical work. This is the narrow-and-deep option.

Structured step-by-step reasoning tool that lets Claude decompose and revise complex problems.

Makes long chains of reasoning inspectable instead of buried in one opaque answer.

Real-time web search, extract, map, and crawl via Tavily's search API, purpose-built for LLM agents.

The search API the agent frameworks standardized on. Generous free tier, boring in the good way.

Retrieves Wikipedia articles, summaries, and sections to ground LLM answers in citable facts.

Unglamorous and constantly useful. Grounding claims against Wikipedia kills a whole class of hallucinations.