Thread Analyzer MCP Server

Scrape and analyze replies from any public Threads post — right from your AI coding agent.

---

What is this?

An MCP server that lets your AI assistant scrape a Threads post URL, then query and analyze the replies — all through natural conversation.

Instead of:

1. Manually open browser
2. Scroll through hundreds of replies
3. Copy-paste into spreadsheet
4. Manually look for patterns

Just say:

"Analyze the replies on this Threads post: https://www.threads.com/@zuck/post/ABC123"

Your AI agent handles the rest.

Features

• Network interception — Captures GraphQL API responses, not fragile CSS selectors that Meta randomizes
• Anti-detection — Randomized scroll delays, stealth browser flags, custom User-Agent
• 4 MCP tools — Scrape, list, search, and get statistics on replies
• Works with any MCP client — Claude Code, Claude Desktop, Cursor, VS Code, Windsurf, and more

MCP Tools

Tool	Description
scrape_thread(url)	Scrape all replies from a public Threads post
get_all_replies()	Return all scraped replies with username and timestamp
search_replies(keyword)	Case-insensitive keyword search across replies
get_reply_stats()	Reply count, top commenters, avg length, time range

Quick Start

Prerequisites

• Python 3.13+
• uv package manager

Installation

git clone https://github.com/ethan-tsai-tsai/thread-analyzer.git
cd thread-analyzer
uv sync
uv run playwright install chromium

Configuration

Add the server to your MCP client config:

{
  "mcpServers": {
    "thread-analyzer": {
      "command": "uv",
      "args": ["run", "--directory", "/absolute/path/to/thread-analyzer", "python", "server.py"]
    }
  }
}

Claude Code

Add to your project's .mcp.json:

{
  "mcpServers": {
    "thread-analyzer": {
      "command": "uv",
      "args": ["run", "--directory", "/absolute/path/to/thread-analyzer", "python", "server.py"]
    }
  }
}

Or run: claude mcp add thread-analyzer -- uv run --directory /absolute/path/to/thread-analyzer python server.py

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "thread-analyzer": {
      "command": "uv",
      "args": ["run", "--directory", "/absolute/path/to/thread-analyzer", "python", "server.py"]
    }
  }
}

Cursor

Go to Cursor Settings > MCP > Add new MCP Server, then add:

{
  "mcpServers": {
    "thread-analyzer": {
      "command": "uv",
      "args": ["run", "--directory", "/absolute/path/to/thread-analyzer", "python", "server.py"]
    }
  }
}

VS Code (Copilot)

Add to .vscode/mcp.json in your workspace:

{
  "servers": {
    "thread-analyzer": {
      "command": "uv",
      "args": ["run", "--directory", "/absolute/path/to/thread-analyzer", "python", "server.py"]
    }
  }
}

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "thread-analyzer": {
      "command": "uv",
      "args": ["run", "--directory", "/absolute/path/to/thread-analyzer", "python", "server.py"]
    }
  }
}

Standalone CLI

You can also use the scraper directly without MCP:

uv run python scraper.py "https://www.threads.com/@user/post/XXXXX"

# Options
uv run python scraper.py "URL" --output custom.csv --max-scrolls 50

How It Works

┌─────────────┐     MCP (stdio)     ┌──────────────┐    Playwright    ┌─────────────┐
│  AI Client  │ ◄──────────────────► │  server.py   │ ◄──────────────► │  Threads.com │
│ (Claude,    │   scrape_thread()    │  (FastMCP)   │   GraphQL API   │  (Meta)      │
│  Cursor...) │   get_all_replies()  │              │   interception  │              │
│             │   search_replies()   │  replies.csv │                 │              │
│             │   get_reply_stats()  │              │                 │              │
└─────────────┘                      └──────────────┘                 └─────────────┘

1. You give your AI assistant a Threads post URL
2. AI calls scrape_thread(url) via MCP
3. Server launches headless Chromium, navigates to the post
4. Playwright intercepts GraphQL network responses containing reply data
5. Server parses replies (username, text, timestamp), saves to CSV
6. AI uses get_all_replies(), search_replies(), get_reply_stats() to analyze

Anti-Detection

Technique	Purpose
Custom User-Agent	Mimics real Chrome browser
navigator.webdriver removal	Hides automation flag
AutomationControlled disabled	Prevents Chromium detection
Randomized scroll delays (1.5-4.5s)	Avoids behavioral fingerprinting
Early stop on idle scrolls	Mimics natural browsing patterns

Limitations

• Public posts only — Cannot access private or restricted posts
• Meta's anti-bot measures — Meta may block headless browsers; if scraping fails, try the standalone CLI in non-headless mode
• GraphQL schema changes — Meta periodically changes their API structure; the parser in scraper.py may need updating

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

License

MIT

---