Maibaamcrawler

@MABAAM

22 days ago

What it does 8 MCP tools that give your AI assistant the ability to research the web, read academic papers, extract YouTube videos, pull tweets, and ingest local files — all through the standard MCP protocol.

Overview Tools Comments

Overview

What it does

8 MCP tools that give your AI assistant the ability to research the web, read academic papers, extract YouTube videos, pull tweets, and ingest local files — all through the standard MCP protocol.

Tools

web_search — 3-tier search cascade (Brave → DuckDuckGo → HTML scraper). Works with or without an API key.
fetch_url — Fetch any URL and convert to clean markdown. SSRF-protected, cached, CAPTCHA-aware.
research — Compound pipeline: search → parallel fetch → summarize → synthesize. Three depth levels.
youtube_essence — Transcript, summary, key points, chapters, and quotes from any YouTube video.
deep_ingest — Extract text from local files: PDF, DOCX, XLSX, PPTX, audio, video, images.
academic_lookup — Resolve DOI, ArXiv ID, or PubMed ID. Fetch full text via institutional access.
twitter_extract — Extract tweets and threads from X.com/Twitter.
vault_status — Check loaded credentials and dependency status.

Key features

Works out of the box — zero configuration needed. Search and fetch work immediately. Add a Brave API key or Ollama for enhanced results.

Credential vault — Configure ~/.mcp-research/vault.yaml to access paywalled content through your university's EZproxy, API keys, or browser cookies. Secrets stay in environment variables, never in plain text.

Token-efficient — Output is capped by default (~4K tokens per fetch) so your AI doesn't waste its context window on raw HTML. All limits are adjustable.

Graceful degradation — Missing optional dependencies don't crash anything. No Ollama? You still get raw results. No yt-dlp? Twitter and YouTube tools tell you how to install it.

Safe — SSRF protection on every request. CAPTCHA detection (Cloudflare, hCaptcha, reCAPTCHA). No eval/exec. No filesystem writes. Cache directories locked to owner-only permissions.

Quick start

pip install mcp-research
mcp-research doctor    # check your setup
Optional extras

pip install 'mcp-research[all]'       # everything
pip install 'mcp-research[youtube]'   # yt-dlp + whisper
pip install 'mcp-research[academic]'  # PDF support
pip install 'mcp-research[twitter]'   # yt-dlp for tweets
pip install 'mcp-research[ingest]'    # PDF, DOCX, XLSX, PPTX, audio
Quick start

npx -y mcp-vfx-parsers
No installation needed — runs directly via npx. Zero external dependencies.

Try in Playground

Server Config

{
  "mcpServers": {
    "research": {
      "command": "uvx",
      "args": [
        "mcp-research"
      ],
      "env": {
        "BRAVE_API_KEY": "",
        "OLLAMA_URL": ""
      }
    }
  }
}

Build with ShipAny.