TradingView MCP vs a Direct Chart Connection

How the TradingView MCP Projects Work

TradingView Desktop is an Electron app, which means it is a Chromium browser underneath. Chromium has a debugging interface, the Chrome DevTools Protocol, that is off by default. Launch the app with the flag --remote-debugging-port=9222 and any program on your machine can connect to that port and read or control what the app is showing.

The TradingView MCP projects are small local servers that do exactly that. Claude Code talks to the server, and the server talks to TradingView Desktop over the debugging port. From Claude's side it looks like a set of tools.

• Requirements listed in the README: TradingView Desktop (a paid TradingView plan), Node.js 18 or later, and Claude Code.
• Setup: clone the repository, run npm install, start TradingView with the debugging port open, then add the server to ~/.claude/.mcp.json.
• What the tools cover: chart state, indicator values, quotes and OHLCV bars; drawing lines, rectangles and text; creating and deleting alerts; writing and compiling Pine Script; screenshots; multi-pane layouts; and replay.

• tradesdontlie/tradingview-mcp on GitHub

  The project's README, including its requirements, tool list and disclaimer.

What Unofficial Means Day to Day

The README is direct about it: the tool is not affiliated with, endorsed by, or associated with TradingView Inc. It also asks users to make sure their use complies with TradingView's terms, which restrict automated data collection, scraping and non-display use of data. That question sits with you, not with the project.

Three practical consequences follow. First, the debugging port has no authentication, so while it is open any program running under your account can drive the app. Second, the tools reach into TradingView's page structure, which changes whenever TradingView ships an update, so a tool can stop working until the maintainer catches up. Third, the data the tools read is still TradingView's data, under TradingView's licence.

None of that makes the project bad. It makes it a clever workaround, and it is worth treating it like one.

What TradingView Itself Offers

TradingView's own answer is the AI Chart Copilot, announced on the TradingView blog as a public beta. It is a browser extension for Chrome and other Chromium browsers that opens a side panel next to your chart. It gives technical breakdowns, manages alerts through conversation, pulls news and fundamentals, and scans your watchlist. TradingView says it is free for now, may limit daily use, and plans to bring it natively into the side panel and the desktop app.

The difference that matters here: the copilot runs TradingView's AI. You cannot point Claude, ChatGPT, Codex or your own agent at it. For a side-by-side of that copilot, see TradingView AI Chart Copilot vs MyLinedChart: Whose AI Actually Works for You?.

• TradingView AI Chart Copilot: public beta

  TradingView's announcement of its own AI side panel.

A Chart Built for a Direct Connection

MyLinedChart takes the opposite approach. Instead of an agent reaching into an app that was never designed for it, the app shares its own chart for your agent to read. There is no debugging port. The connection is a local process that your AI app starts, and it only sees what the chart shares.

Your agent gets nine tools: get_chart_context, get_candles, get_drawings, get_indicators, get_trades, get_provider_status, get_chart_screenshot, get_all_drawing_symbols and propose_chart_changes. The last one is how it draws. Every change is labelled as made by AI, so you can always tell what you drew from what it drew, and any change can be undone. Claude Desktop, Claude Code, Cursor, Codex and Gemini can all connect. Setup steps are on MCP toolkit.

The data comes from a broker or market-data account you already have, connected with your own keys, which stay on your device. MyLinedChart is read-only: it does not place trades.

Side by Side

The two routes answer the same wish, an AI that can see and mark up your chart, from opposite ends.

Which One Fits

If you already live in TradingView Desktop, pay for it, and mainly want help writing Pine Script, a TradingView MCP bridge is a quick experiment. Read its README, close the debugging port when you are not using it, and decide for yourself where you stand on TradingView's terms.

If you want an AI routine you can rely on every morning, data from your own account, and drawings you can export and keep, a chart built for a direct connection is the steadier base. Plenty of traders run both: TradingView for browsing ideas, MyLinedChart for the charts they actually trade from. For what an agent actually reads from a chart, see What an AI Agent Sees on Your Trading Chart.

Is TradingView MCP an official TradingView product?

No. The most popular project, tradesdontlie/tradingview-mcp, states that it is not affiliated with, endorsed by, or associated with TradingView Inc. TradingView's own AI is the AI Chart Copilot browser extension.

Is using a TradingView MCP bridge against TradingView's terms?

The project's README warns that it may conflict with TradingView's terms on automated data collection, scraping and non-display use, and asks each user to check. This is not legal advice: read TradingView's terms for your plan before relying on it.

Can I use ChatGPT or Codex with a TradingView MCP bridge?

The README is written for Claude Code. Other MCP-capable apps may work, but they are not what the project documents. MyLinedChart documents Claude Desktop, Claude Code, Cursor, Codex and Gemini.

Can MyLinedChart read my TradingView charts?

No. MyLinedChart does not connect to TradingView. It charts data from your own broker or market-data account, and TradingView does not export drawings on any plan.