Claude plugged into TradingView
Creators are now wiring Claude — a general LLM — directly into TradingView so the model sits inside trading workflows instead of just explaining charts, signaling a shift toward AI-assisted execution tools (youtube.com). One creator even gave Claude access to 1,500 TradingView scalping strategies, which turns the model into a meta‑layer for sorting, rewriting and testing large strategy libraries rather than inventing trades from scratch (youtube.com). That approach focuses on speeding research and iteration inside systematic workflows, not on promising magical predictive edges, so governance and out‑of‑sample validation become immediate concerns ( ).
The interesting part of this story is not that Claude can “analyze charts.” That was the easy part. The new thing is that creators are wiring Anthropic’s model into TradingView itself, so Claude can read the platform, write Pine Script, run tests, and move through a trader’s normal workflow instead of sitting outside it as a chatbot. Anthropic’s own tooling now supports both direct desktop control through its computer-use feature and external tool connections through the Model Context Protocol, or MCP, which is exactly the kind of plumbing these projects are using (platform.claude.com, claude.com, anthropic.com). That matters because TradingView is not just a charting site. It is where a huge amount of retail and semi-professional systematic trading already lives. TradingView’s own documentation centers Pine Script as the language for building indicators and strategies, and its Strategy Tester is built to simulate those strategies on historical data and then track performance summaries and trade lists (tradingview.com, tradingview.com). Once a model can operate inside that environment, it stops being a commentator on markets and starts becoming a research assistant embedded in the machine room. The clearest example is an open-source bridge called `tradingview-mcp`, published on GitHub by the creator behind Trades Don’t Lie. The repository says it connects Claude Code to a locally running TradingView Desktop app through Chrome DevTools Protocol for chart analysis, Pine Script development, and workflow automation. It also says the tool requires a valid TradingView subscription, runs locally, and depends on undocumented internal TradingView APIs that can break when the desktop app updates (github.com). That last detail is easy to miss, but it explains why this feels less like a polished product launch and more like a fast-moving underground buildout around an existing trading stack. You can see that shift in the videos now circulating around the tool. In one recent demo, a creator shows Claude connected directly to TradingView through an MCP server and asks it to choose indicators, build a Pine Script strategy, backtest it, and optimize settings inside the platform (youtube.com). In another video posted on April 6, 2026, the same creator says he gave Claude access to more than 1,500 “fully backtested and forward-tested” TradingView strategies, then had it analyze them, select among them, and activate or pause bots as part of a live portfolio workflow (youtube.com). That is a different use case from “tell me whether this candle pattern looks bullish.” It turns the model into a meta-layer over an existing strategy library. That is also why the flashy language around “AI trading” is mostly a distraction. The model is not conjuring a secret edge out of thin air. It is sorting, rewriting, parameterizing, and testing a pile of preexisting rules faster than a human can. In the 1,500-strategy demo, the creator explicitly frames Claude as a system for selecting strategies, managing activations, and adapting to market regimes rather than inventing a novel theory of price behavior (youtube.com). The gain is speed. The risk is that speed makes it easier to industrialize bad research. TradingView’s own materials make clear how seductive that can be. A strategy can look impressive in the Strategy Tester because the platform will happily summarize historical performance once a Pine strategy is attached to a chart (tradingview.com). Anthropic’s documentation, meanwhile, warns that computer-use systems need human confirmation for actions with meaningful real-world consequences, including financial transactions, and warns about prompt injection and the dangers of giving a model broad authority over live environments (platform.claude.com). Put those two facts together and the real story comes into focus: these Claude-plus-TradingView setups are making strategy research and execution plumbing much cheaper to automate, right at the point where overfitting, hidden assumptions, and accidental autonomy become expensive. The concrete detail is almost mundane. One of the main bridges doing this work today tells users to pin their TradingView Desktop version, because a routine app update can break the whole connection (github.com).
