Nia-Link MCP Server

念連 (Nia-Link) 🦞

The High-Performance Browser Interface for AI Agents. 「聽見網頁的脈搏，看見資料的靈魂。」 "Hear the pulse of the web, see the soul of the data."

中文 | English

Nia-Link is a Web Neuro-Link Engine purpose-built for AI agents. Natively embracing the Model Context Protocol (MCP), it gives Claude Desktop — or any MCP-compatible AI — 99% reliable web access and automation capabilities.

English Documentation

🚀 Core Advantages

1. Web Neuro-Link

Unlike traditional visual scrapers, Nia-Link listens directly to the browser's CDP (Chrome DevTools Protocol) network pulses. We don't wait for page rendering — we perceive data flow.

2. Action Map Registry

A memory hub with "cached semantics" architecture. Delivers action maps in 0.01s for frequently visited sites, reducing repeated computation costs by 90%.

3. Enterprise-Grade Trust

• 🔒 Snapshot Verified: Every extraction is timestamped and structurally validated.
• 🛡️ Sandbox Isolation: JavaScript executes in isolation, ensuring host safety.
• 📉 Token Optimized: Built-in intelligent filtering saves ~92% context space on average.

4. v0.9 New Features

• 🔄 Website Change Detection (/v1/diff): Track content changes with unified diffs
• 📋 Async Task Queue (/v1/queue/*): Submit background scrape tasks
• ⚡ Rate Limiting: Configurable per-key token bucket rate limiter
• 🧪 Automated Tests: 53 pytest tests covering API, services, and auth

🛠️ Quick Start

Requirements

• Python 3.10+
• Playwright (pip install playwright && playwright install chromium)

Installation

# 1. Install dependencies
pip install -r requirements.txt

# 2. Install browser
playwright install chromium

# 3. Configure environment
cp .env.example .env
# Edit .env to set your API_KEYS

Start the Server

# Development mode
uvicorn app.main:app --reload --port 8000

# Or use Docker
docker compose up -d

Run as MCP Server

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "nia-link": {
      "command": "python",
      "args": ["/path/to/Nia-Link/app/mcp_server.py"]
    }
  }
}

Run Tests

pip install pytest
python -m pytest tests/ -v

📡 API Reference

Health Check

curl http://localhost:8000/health

Scrape a Webpage

curl -X POST http://localhost:8000/v1/scrape \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com",
    "format": "markdown",
    "mode": "fast",
    "extract_actions": true
  }'

Human-like Interaction

curl -X POST http://localhost:8000/v1/interact \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com",
    "actions": [
      {"type": "click", "selector": "#search-btn"},
      {"type": "fill", "selector": "#search-input", "text": "Hello world"},
      {"type": "wait", "ms": 2000}
    ],
    "account_id": "my-session-1"
  }'

Website Change Detection (v0.9)

curl -X POST http://localhost:8000/v1/diff \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com"}'

Async Queue (v0.9)

# Submit task
curl -X POST http://localhost:8000/v1/queue/submit \
  -H "Authorization: Bearer your-api-key" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com"}'

# Poll result
curl http://localhost:8000/v1/queue/{task_id} \
  -H "Authorization: Bearer your-api-key"

MCP Tools

⚙️ Environment Variables

See `.env.example` for all options.

🐳 Docker

docker build -t nia-link .
docker compose up -d
docker compose logs -f

📊 Roadmap

• v0.4: MCP protocol integration
• v0.5: Dynamic gravity particle visualization dashboard