MCP Server - TradingView Chart Image Scraper

🚀 Now with Browser Pooling Optimization for 70-80% Better Concurrent Performance!

This MCP server provides tools to fetch TradingView chart images based on ticker and interval with advanced browser pooling for maximum concurrent performance.

🔥 Performance Optimizations

Browser Pooling Technology

• Production Version: Uses main_optimized.py with browser pooling for maximum concurrent performance
• Performance Improvement: 70-80% faster for concurrent requests
• Concurrent Handling: Supports up to 4 simultaneous requests with pre-initialized browser instances
• Expected Performance:
  • 1 request: ~6-8s (baseline)
  • 2 concurrent: ~3-4s each (60-70% faster)
  • 3 concurrent: ~2.5-3.5s each (70-80% faster)

Version Comparison

• main_optimized.py (Production): Browser pooling, concurrent optimization, performance tracking
• main.py (Legacy): Simple single-browser approach, kept for debugging/fallback

Features

• 🚀 Optimized Chart Image Tool: Fetches direct chart images with browser pooling for maximum concurrent performance
• 📊 Performance Statistics: Built-in performance monitoring and statistics
• 🔄 Browser Pool Management: Pre-initialized browser instances for zero-overhead concurrent requests
• 🎯 Natural Language Prompts: Easy chart requests with interval mapping
• ⚙️ Environment Configuration: Fully configurable via environment variables
• 🔐 TradingView Authentication: Secure session-based authentication
• 💾 Clipboard Capture: Direct base64 image capture for faster performance

Tools

get_tradingview_chart_image

Description: Fetches the direct image URL for a TradingView chart snapshot with optimized concurrent performance.

Performance: This optimized version achieves 70-80% better performance for concurrent requests using browser pooling technology.

Arguments:

• ticker (str): The TradingView ticker symbol (e.g., "BYBIT:BTCUSDT.P", "NASDAQ:AAPL"). Required.
• interval (str): The chart time interval (e.g., '1', '5', '15', '60', '240', 'D', 'W'). Required.

Returns:

• (str): The direct TradingView snapshot image URL (e.g., data:image/png;base64,... or https://s3.tradingview.com/snapshots/...).

Raises:

• Error (MCP type): If the scraper encounters an error. Error codes:
  • 400: Input error (invalid ticker/interval format)
  • 503: Scraper error (failure during the scraping process)
  • 500: Unexpected internal server error

get_performance_stats

Description: Get performance statistics for the optimized TradingView MCP server.

Returns:

• Detailed metrics about request performance, improvement over baseline, and browser pool status

Example Output:

Copy code
🚀 OPTIMIZED TRADINGVIEW MCP SERVER PERFORMANCE STATS
• Total Requests: 12
• Average Time: 3.30s
• Performance Improvement: 70.8%
• Browsers in Pool: 4
• Max Concurrent: 4

Prompts

• Get the {interval} chart for {ticker}
  • Maps common timeframe names (e.g., "1 minute", "5 minute", "1 hour", "daily") to TradingView codes
• Show me the daily TradingView chart for {ticker}
  • Specifically requests the daily ('D') chart
• Fetch TradingView chart image for {ticker} on the {interval} timeframe
  • Comprehensive prompt with timeframe mapping

Setup

1. Create Virtual Environment:

   bashCopy code
   # Navigate to the project directory
   cd tradingview-chart-mcp
   # Create the venv (use python3 if python is not linked)
   python3 -m venv .venv
   

2. Activate Virtual Environment:

   • macOS/Linux:

     bashCopy code
     source .venv/bin/activate
     

   • Windows (Git Bash/WSL):

     bashCopy code
     source .venv/Scripts/activate
     

   • Windows (Command Prompt):

     bashCopy code
     .venv\\Scripts\\activate.bat
     

   • Windows (PowerShell):

     bashCopy code
     .venv\\Scripts\\Activate.ps1
     

     (Note: You might need to adjust PowerShell execution policy: Set-ExecutionPolicy RemoteSigned -Scope CurrentUser)

   Your terminal prompt should now indicate you are in the (.venv).

3. Install Dependencies (inside venv):

   bashCopy code
   pip install -r requirements.txt
   

4. Configure Environment (for Local Testing):

   • Copy .env.example to .env.
   • Fill in your TRADINGVIEW_SESSION_ID and TRADINGVIEW_SESSION_ID_SIGN in the .env file. You can obtain these from your browser's cookies after logging into TradingView.
   • This .env file is used when running the server directly (e.g., python main.py) for local testing.
   • Adjust optional scraper settings (MCP_SCRAPER_HEADLESS, etc.) in .env if needed for local runs.

5. Ensure ChromeDriver: Make sure chromedriver is installed and accessible in your system's PATH, or configure the tview-scraper.py accordingly if it allows specifying a path.

Running the Server

Production (Optimized)

bashCopy code
python main_optimized.py

Legacy (Simple)

bashCopy code
python main.py

Command Line Options

bashCopy code
# HTTP transport with custom port
python main_optimized.py --transport streamable-http --port 8003

# Enable authentication (default for TradingView)
python main_optimized.py --auth

# Adjust concurrency
python main_optimized.py --max-concurrent 6

# Disable browser pooling (fallback to traditional)
python main_optimized.py --disable-pooling

🧪 Performance Testing

Agent-Style Testing

Test the server using the same approach as the production agent:

bashCopy code
cd tests/

# Sequential performance test
python test_mcp_agent_style.py --runs 5 --ticker BYBIT:BTCUSDT.P --interval 240

# Concurrent performance test
python test_mcp_agent_style.py --concurrent 3 --ticker BYBIT:BTCUSDT.P --interval 240

# Test different symbols and timeframes
python test_mcp_agent_style.py --concurrent 4 --ticker NASDAQ:AAPL --interval 15
python test_mcp_agent_style.py --concurrent 2 --ticker BYBIT:ETHUSDT.P --interval D

Expected Results

• Sequential: ~6-8s per request (baseline)
• Concurrent (3x): ~2.5-3.5s per request (70-80% improvement)
• Success Rate: 100% reliability under load
• Throughput: Up to 4 concurrent requests efficiently handled

🔧 Technical Details

Browser Pooling Architecture

• Pre-initialized Browsers: 4 browser instances ready for immediate use
• Thread-Safe Pool: Concurrent access with proper locking
• Async Semaphore: Optimal request limiting
• Performance Tracking: Real-time statistics and monitoring
• Graceful Cleanup: Proper browser lifecycle management

Save Shortcut Feature

The MCP_SCRAPER_USE_SAVE_SHORTCUT feature allows you to capture chart images directly to the clipboard as base64 data URLs:

Benefits:

• Faster Performance: No HTTP requests needed
• More Reliable: No dependency on TradingView's CDN
• Offline Capability: Works once chart is loaded
• Direct Integration: Base64 data URLs for immediate use

Configuration:

bashCopy code
# Enable clipboard image capture (DEFAULT)
MCP_SCRAPER_USE_SAVE_SHORTCUT=True

# Disable and use traditional screenshot links
MCP_SCRAPER_USE_SAVE_SHORTCUT=False

Usage Examples

Available Tools:

• get_tradingview_chart_image(ticker: str, interval: str): Optimized chart fetching with browser pooling

Example Prompts:

• "Get the 15 minute chart for NASDAQ:AAPL"
• "Show me the daily chart for BYBIT:BTCUSDT.P"
• "Fetch TradingView chart image for COINBASE:ETHUSD on the 60 timeframe"

Deactivating the Virtual Environment

When finished:

bashCopy code
deactivate

🔄 Fallback Options

If you encounter issues with the optimized version:

1. Disable Browser Pooling:

   bashCopy code
   python main_optimized.py --disable-pooling
   

2. Use Legacy Version:

   bashCopy code
   python main.py
   

3. Debug Mode:

   bashCopy code
   python main_optimized.py --log-level DEBUG
   

📊 Performance Monitoring

The optimized server includes built-in performance monitoring:

• Request success rates
• Average response times
• Performance improvement metrics
• Browser pool utilization
• Concurrent request handling statistics

Access these metrics via the get_performance_stats tool or through the server logs.

🔌 Using with MCP Clients (Claude Desktop / Cursor)

This server supports two ways of providing configuration:

1. Via .env file (for local testing): When running python main.py directly, the server will load credentials and settings from a .env file in the project directory.
2. Via Client Environment Variables (Recommended for Integration): When run by an MCP client (like Claude/Cursor), you should configure the client to inject the required environment variables directly. These will override any values found in a .env file.

Claude Desktop

1. Open your Claude Desktop configuration file:

   • Windows: %APPDATA%\\Claude\\claude_desktop_config.json
   • macOS: ~/Library/Application\ Support/Claude/claude_desktop_config.json

2. Add or merge the following within the mcpServers object. Provide your credentials in the env block:

   jsonCopy code
   {
     "mcpServers": {
       "tradingview-chart-mcp": {
         "command": "/absolute/path/to/your/tradingview-chart-mcp/.venv/bin/python3",
         "args": ["/absolute/path/to/your/tradingview-chart-mcp/main.py"],
         "env": {
           "TRADINGVIEW_SESSION_ID": "YOUR_SESSION_ID_HERE",
           "TRADINGVIEW_SESSION_ID_SIGN": "YOUR_SESSION_ID_SIGN_HERE"
           // Optional: Add MCP_SCRAPER_* variables here too if needed
           // "MCP_SCRAPER_HEADLESS": "False"
         }
       }
       // ... other servers if any ...
     }
   }
   

3. Replace the placeholder paths (command, args) with your actual absolute paths.

4. Replace YOUR_SESSION_ID_HERE and YOUR_SESSION_ID_SIGN_HERE with your actual TradingView credentials.

5. Restart Claude Desktop.

Cursor

1. Go to: Settings -> Cursor Settings -> MCP -> Edit User MCP Config (~/.cursor/mcp.json).

2. Add or merge the following within the mcpServers object. Provide your credentials in the env block:

   jsonCopy code
   {
     "mcpServers": {
       "tradingview-chart-mcp": {
         "command": "/absolute/path/to/your/tradingview-chart-mcp/.venv/bin/python3",
         "args": ["/absolute/path/to/your/tradingview-chart-mcp/main.py"],
         "env": {
           "TRADINGVIEW_SESSION_ID": "YOUR_SESSION_ID_HERE",
           "TRADINGVIEW_SESSION_ID_SIGN": "YOUR_SESSION_ID_SIGN_HERE"
           // Optional: Add MCP_SCRAPER_* variables here too if needed
           // "MCP_SCRAPER_HEADLESS": "False"
         }
       }
       // ... other servers if any ...
     }
   }
   

3. Replace the placeholder paths (command, args) with your actual absolute paths.

4. Replace YOUR_SESSION_ID_HERE and YOUR_SESSION_ID_SIGN_HERE with your actual TradingView credentials.

5. Restart Cursor.

Installing via Smithery

To install TradingView Chart Image Scraper for Claude Desktop automatically via Smithery:

bashCopy code
npx -y @smithery/cli install @ertugrul59/tradingview-chart-mcp --client claude

Configuration

Environment Variables

The following environment variables can be set to configure the scraper:

• TRADINGVIEW_SESSION_ID: Your TradingView session ID (required)
• TRADINGVIEW_SESSION_ID_SIGN: Your TradingView session ID signature (required)
• MCP_SCRAPER_HEADLESS: Run browser in headless mode (default: True)
• MCP_SCRAPER_WINDOW_WIDTH: Browser window width (default: 1400)
• MCP_SCRAPER_WINDOW_HEIGHT: Browser window height (default: 1400)
• MCP_SCRAPER_USE_SAVE_SHORTCUT: Use clipboard image capture instead of screenshot links (default: True)
• MCP_SCRAPER_CHART_PAGE_ID: Custom chart page ID (optional)

Save Shortcut Feature

The MCP_SCRAPER_USE_SAVE_SHORTCUT feature allows you to capture chart images directly to the clipboard as base64 data URLs instead of getting screenshot links. This eliminates the need to download images from URLs.

Benefits:

• Faster chart capture (no HTTP requests needed)
• More reliable (no dependency on TradingView's CDN)
• Works offline once the chart is loaded
• Direct base64 data URLs for immediate use

How it works:

• When enabled (True): Uses Shift+Ctrl+S (or Shift+Cmd+S on Mac) to capture chart image directly to clipboard
• When disabled (False): Uses traditional Alt+S to get screenshot links, then converts to image URLs

Configuration:

bashCopy code
# Enable clipboard image capture (DEFAULT)
MCP_SCRAPER_USE_SAVE_SHORTCUT=True

# Disable and use traditional screenshot links
MCP_SCRAPER_USE_SAVE_SHORTCUT=False