wtf is tradingview-mcp?

TradingView MCP Bridge connects the Claude Code AI assistant to your own locally running TradingView Desktop app, letting an AI help you read charts, write Pine Script indicators, and automate parts of your chart workflow. It is not made by or connected to TradingView Inc, and it does not bypass any paywall: you still need your own paid TradingView subscription and the Desktop app already installed and running. The tool works by talking to TradingView Desktop through the Chrome DevTools Protocol, a standard debugging interface built into Chromium and Electron based apps like TradingView Desktop, VS Code, Slack, and Discord. That debug connection is off by default and only turns on if you explicitly launch TradingView with a special flag, so nothing is exposed without you choosing to enable it. All communication happens on your own computer. No TradingView data is sent to any external server by this tool, and it cannot place real trades, since it only interacts with the chart interface itself. Once connected, the AI assistant gains a wide set of abilities over your chart: writing, compiling, and debugging Pine Script scripts, switching symbols and timeframes, drawing trend lines and annotations, creating and managing price alerts, replaying historical price bars, taking chart screenshots for visual analysis, arranging multi chart layouts, and streaming live indicator or price data for local monitoring scripts. Every one of these actions is also available as a standalone command line tool called tv, so scripts and other programs can use the same features outside of Claude Code, with JSON output designed for piping into tools like jq. Setup involves cloning the repository, installing its Node.js dependencies, launching TradingView Desktop with the debug port enabled using a provided script for your operating system, and registering the tool in Claude Code's MCP configuration file. The project frames itself partly as a research effort into how AI agents can work with financial desktop applications, and includes a research document discussing open questions and known failure modes, rather than positioning itself as a trading bot.