How To Connect Claude to TradingView

ELI5/TLDR

Someone built a tool that lets Claude read TradingView charts in real time — not by looking at screenshots, but by reading the actual data underneath. You talk to Claude in plain English, it manipulates your charts, writes trading strategies as code, and gives you a daily briefing across all your assets. Lewis Jackson took the original project, fixed some rough edges, added a “morning brief” feature, and is giving it away free.

The Full Story

The Problem With AI and Charts

Every previous attempt to marry AI with trading charts has had the same flaw. You screenshot the chart, paste it into an AI, and by the time the AI responds, the chart has moved. The AI was also just squinting at pixels — interpreting an image, not reading data. A slightly blurry screenshot could mean the difference between “buy” and “don’t buy.” Fine details matter in trading, and pixel-guessing is a poor way to capture them.

What Makes This Different

The tool is called TradingView MCP, originally created by a developer called Trades Don’t Lie. MCP stands for Model Context Protocol, but Jackson waves that away — the point is what it does, not what the acronym means.

Instead of looking at an image, Claude reads the live code behind TradingView’s interface. Every website is just code rendered visually. If you right-click any webpage and hit “Inspect,” you see the raw data underneath. TradingView MCP essentially gives Claude permanent access to that inspect panel, reading the actual numerical values of every candle, every wick, every indicator — updated in real time, even on a 1-second chart.

“Claude isn’t looking at an image. It’s actually looking at the code and the values on the screen.”

The difference is like asking someone to estimate a building’s height from a photo versus handing them the architectural blueprints. One is a guess. The other is the answer.

How It Works Under the Hood

TradingView’s desktop app exposes something called CDP (Chrome DevTools Protocol). This is the same interface that browser developer tools use. The MCP server connects Claude to this protocol, so Claude can both read what’s on screen and send commands back — changing charts, adding indicators, injecting Pine Script code.

You talk to Claude in your terminal. Claude interprets your natural language, translates it into the right commands, and TradingView responds. Ask it to switch to the Bitcoin weekly chart, and it just does it. Ask it to remove the volume indicator, and it vanishes.

The Setup

Jackson skips the manual installation (which involves cloning a GitHub repo and configuring files by hand) and offers a one-shot setup prompt. You paste it into Claude Code, answer a couple of yes/no questions, and the system configures itself. The main requirement is having TradingView Desktop installed and launched with CDP enabled.

The setup creates an mcp.json configuration file and a rules.json file where your trading preferences live.

Building a Strategy by Talking

This is where it gets interesting. Jackson doesn’t have a trading strategy of his own, so he simply asks Claude to research prominent public traders, figure out what indicators they use, and implement that as a strategy.

Claude comes back with a combination of two well-known approaches: the Van de Pop setup (good for Bitcoin swing trading on the weekly) and the Tone Vays method, using EMAs at 21, 50, and 200, plus RSI at 14 and MACD. It writes these into the rules.json file, then applies the indicators directly onto the chart.

Then Jackson asks Claude to convert the whole strategy into a Pine Script — TradingView’s proprietary scripting language that lets you create custom indicators and backtest strategies against historical data. Claude writes the script, applies it to the chart, notices errors in the code (because it can read TradingView’s error messages in real time), and fixes them autonomously.

“Because Claude code is watching what’s on the screen from a code level, it can see that there is one of two problems and it will go in there and fix those issues and it will keep doing it and keep fixing them until it’s all resolved.”

The debugging loop — write code, see error, fix code, repeat — happens without any human intervention. For anyone who’s manually debugged Pine Script, this is noteworthy.

The Watchlist and Morning Brief

The original tool required you to manually ask Claude to analyze each chart, one at a time. Jackson found this tedious, so he added two things:

1. A watchlist in rules.json — you tell Claude which assets you care about (Bitcoin, Ethereum, Solana, XRP, Chainlink, Pepe, whatever), and it stores them.

2. A morning_brief command — type it once, and Claude cycles through every asset on your watchlist, applies your strategy, and delivers a summary. Bitcoin’s bearish. Ethereum’s bearish. XRP is neutral. Here’s what to do. One command, all your assets.

You could even schedule this to run every 10 minutes if you wanted a persistent feed, though Jackson notes that might get annoying.

The Copy-a-Trader Idea

Jackson floats a more ambitious use case: scrape every video transcript from a technical analyst you admire — someone like CoinsKid or Blockchain Backer — feed those transcripts to Claude, have it extract their methodology, and convert that into a Pine Script strategy on your charts.

“What happens in the end is you end up looking at a chart the same way that CoinsKid does.”

The idea is to separate the strategy from the emotion. You do the strategic thinking once (or borrow someone else’s), encode it, and then just follow the signals. No second-guessing in the moment.

Claude’s Take

The underlying technology here is real and genuinely useful. MCP (Model Context Protocol) connecting Claude to TradingView via CDP is a legitimate integration pattern, and reading structured data instead of screenshots is an obviously better approach. That part checks out.

What deserves some skepticism is the framing. The video implies that connecting an AI to a chart is most of the battle, when in reality, the hard part was always the strategy itself. A perfectly executed bad strategy is still a bad strategy. Jackson freely admits he doesn’t have a trading strategy and asks Claude to just research one — but “research prominent traders and implement their strategy” is doing a lot of heavy lifting there. Claude isn’t backtesting these strategies rigorously or evaluating their statistical edge. It’s pattern-matching public information about what indicators someone claims to use.

The “copy a trader” idea is appealing in theory but papering over a real problem. Knowing that someone uses a 21-week EMA doesn’t tell you how they actually make decisions — the judgment calls, the exceptions, the context they factor in that never makes it into a YouTube video. You’d be copying the skeleton, not the brain.

The morning brief feature is the most practically valuable addition. Automating a repetitive workflow — cycling through a watchlist and applying consistent criteria — is exactly what these tools are good at. That’s a genuine time-saver with a clear use case.

Jackson also casually mentions his “10X dashboard” and a strategy to turn 50,000 pounds into 500,000 in a year, with links to paid products in the description. The video itself is free, which is nice. But the broader channel is clearly a funnel toward selling trading tools and alerts, which is worth keeping in mind when evaluating the enthusiasm level.

The tool is real. The integration is clever. The idea that it will make you “a ton of money” is doing the thing that trading content always does — conflating better tooling with better outcomes. A sharper knife doesn’t make you a chef.