Claude Execute — Automated Crypto Trading

A lightweight automation layer that reads your TradingView chart, evaluates your strategy, and places trades on your exchange — with strict safety checks and full tax-ready logging.

---

What This Does

Five things you get from this setup:

1. Claude connected to your exchange — reads your TradingView chart and executes trades on your exchange automatically
2. A safety check — every condition in your strategy must pass before a single trade goes through
3. 24/7 cloud execution — deploy to Railway and it runs on a schedule, even when your laptop is closed
4. Automatic tax accounting — every trade logged to trades.csv with date, price, fees, and net amount, ready for your accountant
5. Fully yours — no email, no course, no upsell. Everything is in this repo.

---

The One-Shot Prompt

This is the thing you paste. Open Claude Code in this directory, paste the entire contents of prompts/02-one-shot-trade.md, and Claude will do the rest.

Here's what it does when you run it:

Step	What Claude does
1	Reads your rules.json strategy
2	Pulls live price + indicator data from TradingView
3	Calculates MACD from raw candle data
4	Evaluates market bias (bullish / bearish / neutral)
4b	Checks trade limits — daily cap and max trade size
5	Runs the safety check — every entry condition checked
6	Executes the trade via your exchange if all conditions pass
7	Logs the trade to trades.csv — date, price, fees, net amount (tax-ready)
8	Saves full decision log to safety-check-log.json

If anything fails the safety check, it stops and tells you exactly which condition failed and the actual values. No trade goes through unless everything lines up.

---

Getting Started

Step 1 — Paste the one-shot prompt into Claude Code

Copy the entire contents of prompts/02-one-shot-trade.md and paste it into your Claude Code terminal.

That's it. Claude acts as your onboarding agent — it clones the repo, walks you through connecting your exchange, sets your trading preferences, connects TradingView, optionally builds a strategy from a YouTube channel, deploys to Railway, and runs the bot for the first time. Every step is interactive. It pauses when it needs something from you and handles everything else automatically.

---

What's Happening Under the Hood

For anyone who wants to understand the steps manually, or troubleshoot a specific part:

Prerequisites

• TradingView MCP must already be set up — setup walkthrough video
• Claude Code installed and running
• An exchange account with API access (Binance, BitGet, Bybit, OKX, Coinbase Advanced, Kraken, KuCoin, Gate.io, MEXC, or Bitfinex)
• Node.js 18+ — check with node --version

---

Clone the repo

Mac / Linux:

git clone https://github.com/vaughanf1/claude-execute
cd claude-execute

Windows:

git clone https://github.com/vaughanf1/claude-execute
cd claude-execute

---

Add your exchange API credentials

Mac / Linux:

cp .env.example .env

Windows:

Copy-Item .env.example .env

Open .env and fill in:

BITGET_API_KEY=your_api_key_here
BITGET_SECRET_KEY=your_secret_key_here
BITGET_PASSPHRASE=your_passphrase_here
PORTFOLIO_VALUE_USD=1000
MAX_TRADE_SIZE_USD=100
MAX_TRADES_PER_DAY=3

The BITGET_* variable names are kept for compatibility with the bot code — fill them in with credentials from whichever exchange you use. If your exchange doesn't use a passphrase (e.g. Binance), leave that field blank.

Getting your API key:

Step-by-step guides for all supported exchanges:

Exchange	Guide
BitGet	docs/exchanges/bitget.md
Binance	docs/exchanges/binance.md
Bybit	docs/exchanges/bybit.md
OKX	docs/exchanges/okx.md
Coinbase Advanced	docs/exchanges/coinbase.md
Kraken	docs/exchanges/kraken.md
KuCoin	docs/exchanges/kucoin.md
Gate.io	docs/exchanges/gateio.md
MEXC	docs/exchanges/mexc.md
Bitfinex	docs/exchanges/bitfinex.md

Two rules that apply to every exchange — withdrawals OFF, IP whitelist ON.

---

Launch TradingView and connect the MCP

Mac:

tv_launch
tv_health_check

Windows: See docs/setup-windows.md

Linux: See docs/setup-linux.md

Verify with tv_health_check — should return cdp_connected: true.

---

Run the bot manually

node bot.js

---

Terminal Dashboard

A Matrix-style live dashboard for viewing every decision the bot has made — executed trades, blocked trades, and the exact conditions that passed or failed.

npm run dashboard

Then open http://localhost:3737.

What you get:

• Stats strip — total decisions, trades taken, blocked, today's count vs daily cap, total volume, estimated fees
• Executed Trades — every order that went through (paper or live), with timestamp, price, size, mode, order ID
• Decision Feed — every safety-check run, color-coded PASS / BLOCK, filterable
• Decision Detail — click any row to inspect every condition (✓ pass / ✗ fail), the indicator values at decision time, and the trade size/limits

Zero dependencies, reads safety-check-log.json and trades.csv directly. Auto-refreshes every 5 seconds.

Want to see the dashboard populated before running the bot? Seed it with demo data:

node dashboard/seed-demo.cjs        # 12 sample decisions (3 PASS, 9 BLOCK)
node dashboard/seed-demo.cjs reset  # remove demo data

---

Deploy to Railway (Run in the Cloud 24/7)

The local setup runs when your laptop is open. Railway lets the bot check for setups around the clock — even while you sleep.

Note: Cloud mode pulls candle data directly from Binance's free market API instead of TradingView. No TradingView Desktop needed in the cloud. The strategy logic and safety check are identical.

1. Deploy

npm install -g @railway/cli
railway login
railway init
railway up

2. Set your environment variables in Railway

Go to your Railway project → Variables and add everything from .env.example:

Variable	Example
BITGET_API_KEY	your key
BITGET_SECRET_KEY	your secret
BITGET_PASSPHRASE	your passphrase (blank if not applicable)
PORTFOLIO_VALUE_USD	1000
MAX_TRADE_SIZE_USD	100
MAX_TRADES_PER_DAY	3
PAPER_TRADING	true (set to false when ready)
SYMBOL	BTCUSDT
TIMEFRAME	4H

3. Set a cron schedule

In Railway → Settings → Cron Schedule, set how often the bot runs. Recommended:

Timeframe	Schedule	What it means
4H chart	0 */4 * * *	Every 4 hours
1D chart	0 9 * * *	Once a day at 9am UTC
1H chart	0 * * * *	Every hour

4. Start in paper trading mode

PAPER_TRADING=true logs every decision but never places real orders. Watch a few days of paper trades, confirm the logic matches what you expect, then flip it to false.

---

Build Your Own Strategy (Optional)

The example rules.json uses a VWAP + RSI(3) + EMA(8) scalping strategy. To build one from any trader's public videos:

1. Go to Apify and search the actor store for YouTube Transcript Scraper — takes about 30 seconds per channel
2. Paste the output into prompts/01-extract-strategy.md
3. Run that prompt in Claude Code — it generates a rules.json tailored to that trader's methodology

---

Files

File	What it does
rules.json	Your strategy — indicators, entry rules, risk rules
.env	Your exchange credentials (gitignored — never commits)
prompts/01-extract-strategy.md	Build rules.json from trader transcripts
prompts/02-one-shot-trade.md	The one-shot prompt — paste this to trade
safety-check-log.json	Auto-generated log of every trade decision
trades.csv	Tax-ready trade record — auto-written on every execution
docs/setup-windows.md	Windows-specific MCP setup
docs/setup-linux.md	Linux-specific MCP setup

---

Tax Accounting

Every trade the bot places is automatically written to trades.csv with the columns your accountant needs:

Column	Description
Date	ISO date of the trade
Time	UTC time
Exchange	Exchange name
Symbol	e.g. BTCUSDT
Side	Buy / Sell
Quantity	Units traded
Price	Price per unit at execution
Total USD	Gross trade value
Fee (est.)	Estimated exchange fee
Net Amount	Total USD minus fee
Order ID	Exchange reference
Mode	Paper / Live

At tax time: open the file, hand it to your accountant, or import it directly into your accounting software. Nothing to reconstruct.

For a quick summary of your trading activity, run:

node bot.js --tax-summary

This prints total trades, volume, and fees paid.

---

Safety

The safety check conditions are not fixed — they come directly from your rules.json. If you build a strategy from a YouTube trader's transcripts using the Apify prompt, your safety check will reflect that trader's entry logic. If you use the example strategy, it reflects those conditions. They're yours, not a generic filter.

Every condition in your entry_rules must pass before a trade goes through. One fails — nothing happens. The bot tells you exactly which condition failed and the actual value it saw.

Additional guardrails that apply regardless of strategy:

• Maximum trade size capped at MAX_TRADE_SIZE_USD in .env
• Maximum trades per day capped at MAX_TRADES_PER_DAY in .env
• Position sizing calculated from your portfolio value — max 1% risk per trade
• Every decision logged to safety-check-log.json with exact indicator values
• Every executed trade recorded in trades.csv for accounting

This is not financial advice. Build your strategy properly. Run the backtest. Paper trade before going live. Never put in more than you can afford to lose.

---

Resources

• Apify — search actor store for "YouTube Transcript Scraper"
• Railway — for 24/7 cloud execution
• TradingView — charting + MCP source