LLMs plugged into TradingView
Creators are no longer just asking AI for market commentary — they’re wiring large language models directly into charting tools so the model becomes the interface for strategy research and code work. Two recent videos show users connecting Claude to TradingView and using the model to inspect scripts, which shifts AI’s role from adviser to workflow orchestrator and suggests demand for native AI chart assistants ( ).
The old way to use AI in trading was simple. You pasted a chart screenshot into a chatbot and asked what it saw. The new way is stranger and more powerful. Traders are now wiring large language models directly into TradingView, so the model can inspect the charting environment itself, read scripts, and act more like an operator than an oracle. In one video posted on April 4, a creator walks through connecting Claude Code to TradingView with a local MCP server and Chrome-style remote debugging, then asks the model to read the chart and build indicators in Pine Script. In another, posted on April 6, a creator says he gave Claude access to more than 1,500 TradingView strategies and had it choose which bots to activate or pause (youtube.com, youtube.com). That matters because TradingView is not a niche toy. It is one of the main workbenches for retail traders, and Pine Script is the language that lets them turn chart ideas into indicators, alerts, and backtests inside the platform (tradingview.com). Once a model can move through that environment, inspect page state, and write or revise Pine code in context, the chat window stops being a place for commentary. It becomes the front door to the workflow. The enabling technology did not come from TradingView. It came from the recent push to make models use ordinary software the way people do. Anthropic described this shift in late 2024 when it introduced Claude’s “computer use” capability, which lets the model look at screenshots, move a cursor, click, and type inside existing applications instead of relying only on custom APIs (anthropic.com). The creators now stitching Claude into TradingView are taking that general capability and aiming it at one of the most detail-heavy interfaces on the consumer web. The setup is still improvised. The April 4 tutorial does not show an official Claude-TradingView product. It shows a stack of parts: Claude Code, a GitHub-hosted TradingView MCP project, Node.js, and a TradingView desktop app launched with a remote debugging port so Claude can interrogate what is on screen (youtube.com, github.com). That is exactly why the videos are interesting. When users are willing to bolt together developer tools just to let a model inspect chart state and write strategy code, they are signaling demand before the platform owner has fully packaged the feature. TradingView has clearly noticed. On April 2, the company announced a public beta for “TradingView AI Chart Copilot,” a browser-side assistant that lives in a side panel next to charts. TradingView says it can explain technical setups, manage alerts through conversation, pull in news and fundamentals, and scan for trade ideas. The company also says the beta may have daily usage limits and frames the product as the first step toward bringing TradingView AI natively into the side panel across browsers and the desktop app (tradingview.com). That timing makes the broader shift easier to see. Independent creators are already pushing LLMs past market punditry and into direct tool control. TradingView is moving in the same direction, but with a narrower first product that helps interpret charts and manage platform actions rather than write full Pine strategies. The gap between those two things is now the whole story. One side is a rough agentic hack built from MCP servers and debug ports. The other is a polished assistant in a browser panel. Both start from the same premise: the chat box should sit beside the chart, not outside it (tradingview.com, youtube.com).
