Scott’s List · AI research tool review

What it is

Claude Code is Anthropic’s AI assistant for people who work with files and code, and you don’t have to be a programmer to use it. It runs on your own computer, in a terminal window or the desktop app, and it can read the files you point it at, write and run scripts, and connect to other tools. For a trader that means it can open a spreadsheet of trades, pull price bars, write an indicator, and hand back a table, a chart, or a plain answer. You talk to it in ordinary sentences. It does the typing.

The facts at a glance

• Made by Anthropic, the company behind the Claude models. Runs in a terminal, the Claude desktop app, or inside VS Code and JetBrains editors.
• Included with the paid Claude subscriptions (Pro and Max) at the time of writing, or billed by usage through Anthropic’s API. Rate limits depend on the plan.
• Works on your machine: it reads and edits files you allow, runs scripts, and can connect to outside tools such as TradingView through what Anthropic calls MCP connectors.
• Has no market data of its own. It works on what you give it: your journal, your exported bars, your rules.
• It can be confidently wrong. Every number it gives me gets checked against the source before I act on it.

Details are as published at the time of writing and change without notice. Read the current terms on the company’s own site before you pay for anything.

How it fits the VWAP Wave System

The system gives me four setups, three conditions, and a grade. Every one of those is a claim about how the market behaves, and I spent twenty-five years accepting claims like that on faith. Claude Code is the first tool I’ve had that lets me put a count behind them without a week of spreadsheet work. It’s a confirmation tool in the truest sense: it confirms or denies what I believe, after the close, with my own data.

Three ways it lives in my stack:

• The journal math. My journal grades every trade twice, once for the setup and once for the execution. Claude Code reads that file and answers the questions I’d otherwise guess at. How do my A-grade VWAP Bounces do on balanced days versus discovery days? How much of my loss column is trades with no setup name at all? The answers changed my trading more than any indicator, and they’re in the execution column lesson.
• The chart. The Pine scripts on my TradingView layout were written and debugged with it, including the session VWAP with its deviation bands, the Initial Balance, and the value-area levels I use for confluence. I describe what I want in plain English, it writes the script, compiles it against TradingView, and fixes its own errors. The textbook charts in the free course on this site were rendered the same way, from real ES bars.
• The checklist and the site. The pre-market and trade checklist I run every session started as a conversation with it, and this site is built and maintained with it. When I say the evaluator’s desk on Scott’s List grades things by evidence, this is part of how the evidence gets counted.

What it doesn’t do is anything during the session. It has never seen my order entry and it never will. That’s a rule, not a limitation. The Cliff Young lesson is about not inheriting the industry’s truths; this is the tool that lets me test them instead.

Scott’s notes

Two warnings I’d give anyone starting with it. First, it will hand you a clean-looking statistic from a data set of eleven trades and not sound any less sure of itself than it does with eleven hundred. Sample size is your job. Second, it’s happy to build you a strategy that backtests beautifully because it quietly fit the rules to the past. I use it to check rules I already trade, not to invent new ones.

What surprised me is how much of the value is in the questions it makes me ask. Once the counting is cheap, you stop protecting your beliefs, because finding out is easier than defending them.

Verdict