From Pre-Market Scan to Telegram Signal: A Full Trading Pipeline With Claude

The host opens with a warning: machine intelligence alone turns nobody profitable; real desk experience comes first and the models only sharpen an existing edge. He presents himself as non-technical, which frames the whole piece as a trader bending tools to a workflow rather than a coder showing off.

The rig needs three pieces: the Claude Code desktop app, the TradingView desktop app, and a community MCP bridge between them, installed from copy-paste prompts on his blog. One wrinkle: the link worked on his Mac while the Windows attempt failed on a desktop-file mismatch.

The price of admission is two subscriptions: at least the 20-dollar monthly Claude tier, he runs the 100-dollar Max tier himself, plus a paid TradingView plan, with the cheapest Essential tier around 13 to 15 dollars a month doing the job.

A health-check prompt confirms the live link, the chart is verified against AMD and dropped to the five-minute frame, and the first tricks are auto-drawn levels for Intel: recent highs, pre-market extremes, 52-week bands and all-time peaks, each styled and labeled.

Then comes the key constraint: the bridge cannot sweep the whole market in one go, it only reads a basket you hand it. His fix is a two-stage design where scanner A collects the day's gappers and scanner B filters that shortlist through strategy rules.

Scanner A hunts pre-market movers each morning on three cutoffs: a gap above 5 percent, a share price over 3 dollars and pre-market volume past 50 thousand. Quotes come from Yahoo Finance and catalysts from Benzinga, printed as ticker, price, gap and a one-line catalyst, then archived into a dated JSON file.

Scanner B encodes his trend-join-long entry screen across five gates: only after 10 a.m., price above the prior daily high, prior close above the 200-day average, price above the pre-market high and above the day's high. A clean sweep fires an entry, and this stage depends on the desktop app staying open with the chart tab live.

Claude ports the screen into Pine Script and replays MU on the five-minute chart: 14 trades netting 1,200 dollars in the demo with a 2.48 profit factor. The host flags the catch himself: a single ticker, and names like AMD sitting inside a powerful run, so nothing here generalizes yet.

The replay widens to the mega-cap basket on the 15-minute frame and stays decent while turning highly selective, with Tesla printing no trades at all. Strict gates mean fewer tickets by construction, and he frames that as a feature for win-rate hunters.

Hitting Pine Script's ticker ceiling, he has Claude rewrite the same screen in Python and sweep a 30-name watchlist over 30 days: a 10,000-dollar book at a 54 percent win rate, with Meta and Robinhood silent. The Python loop runs visibly faster than the chart-by-chart replay.

Both scanners get scheduled: scanner A every half hour from 8:30 a.m. to 2 p.m., the strategy screen five minutes behind it from 10 a.m. to 2:30 p.m., twelve cycles a day firing on their own.

Each cycle lands on his phone through Telegram in a fixed layout, gappers plus catalysts first, strategy candidates after the open. The video walks through BotFather, token handoff and format locking step by step.

The bonus act is TradingView's own beta assistant, a browser side panel that replays the watchlist across 90 days and rebuilds the gap scan on demand. It cannot push runs to Telegram, so mornings stay manual for now.

He closes where automation honestly stops: Pine Script replays but never trades, so he has moved the screened strategy onto live rails through the Interactive Brokers API and promises the full wiring in next week's video.