Personal finance · checked 2026-10-19Can I vibecode TradingWizard?
KINDA· weekend project
The loop is easy to describe: fetch candles for a watchlist, compute indicators, ask a model for a Buy, Sell or Wait read with levels and a reason, and alert when it changes. An agent builds a personal version in a sitting. The hard part is data, not code: crypto candles are free from exchanges, while real-time stocks, forex and futures are licensed. A home-built read also has no track record behind it, so treat what you build as a research notebook, not advice; the product itself says it provides analysis and places no orders.
$40/moprice
$480you'd save a year
weekendbuild time
—people replaced it
What you need on hand.
DifficultyIntermediate · Some system setup or keys.
TimeA weekend; data sources and alert setup take the longest
- Claude Code, Codex or CursorAny of them can run the prompt.
- Node.js 22better-sqlite3 compiles a native module, so build tools must be installed.
- An LLM API key or OllamaAnthropic or OpenAI for the best reads; Ollama keeps everything on your machine.
- A data sourceExchange public endpoints need no key; stocks come from an unofficial, delayed feed unless you pay a provider.
- An alert channelAn ntfy topic or a Telegram bot, both free.
Before you buildTradingWizard has no free plan, but TradingView's free plan covers charts and a few price alerts if that is all you need.
- 1. Install one agent: Claude Code, Codex, Cursor.
- 2. Make an empty folder and open it in the agent.
- 3. Paste the prompt below, press enter, and answer its questions as it builds.
The prompt
Build me a personal market watch to replace TradingWizard for a short list of assets. Requirements: - Node.js 22 + TypeScript: a Fastify API and a React UI built with Vite. `npm start` serves http://localhost:5400 and runs the scanner in the same process. - Storage: better-sqlite3 in ./data/watch.db: watchlist (symbol, market crypto or stock, timeframe 1h / 4h / 1d), candles, reads (time, label Buy / Sell / Wait / Avoid, entry, stop, target, a two-sentence reason, model name) and alerts sent. Keys and settings are in .env. - Data behind one CandleSource interface: crypto candles from a public exchange endpoint through `ccxt` (no key); stocks through `yahoo-finance2` (unofficial and possibly delayed, so say so in the UI). If the newest candle is older than twice the timeframe, skip the scan and show "data stale". - Indicators computed in code with `trading-signals`: EMA 20/50/200, RSI 14, ATR 14, MACD, volume against its 20-bar average and the nearest swing high and low. The model receives this table, never raw prices. - Reads: per item on its own schedule (`node-cron`; crypto around the clock, stocks while the exchange is open), send the table to an LLM from .env (ANTHROPIC_API_KEY, OPENAI_API_KEY or OLLAMA_URL) and require JSON validated with `zod`. Replace a read with Wait when its stop is on the wrong side of the entry or its levels sit more than 3 ATR from the last close. - Paper ledger: a Buy or Sell read opens a simulated position that closes at its stop or target using later candles; show closed trades, wins and losses per symbol, losses included. Simulated only: no broker or exchange fields exist anywhere in the app. - Alerts when a label changes, through ntfy (NTFY_TOPIC) or a Telegram bot (`grammy`). Every screen and alert carries "Analysis, not financial advice. Simulated trades only." - UI: watchlist table with label, levels, reason and age; a symbol page with a candle chart (`lightweight-charts`) and its history. No accounts, no telemetry, local-first; outbound calls go to the data source, your LLM and your alert channel. - Out of scope: real orders, futures and forex feeds, a mobile app, copy trading, performance claims. - README: setup, data caveats and terms of the unofficial feed, the cost of one scan, and a warning that nothing here has a proven edge: test before relying on any read.
Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.
Then keep going
Next prompts, same folder.
The first prompt builds the core. Run these one at a time in the same folder to grow it into your own TradingWizard.
Add a backtest to the market watch app in this folder. Requirements: - A command `watch backtest <symbol> --from <date> --to <date>` that replays stored candles bar by bar, using only data up to each bar, and applies the same indicator code. - Use a rule-based stand-in for the LLM (the label comes from EMA and RSI thresholds in config), not live model calls, and say so in the output. - Report trade count, win and loss counts, average win and loss in ATR units, maximum drawdown and the buy-and-hold change over the same period, with fees and slippage set in config. - Print a warning that past results do not predict future ones. - Update the README.
Add a daily digest to the market watch app in this folder. Requirements: - At a time in the timezone from .env, send one message through the existing alert channel: each watchlist item with its current label, how long it has held it, and any change in the last 24 hours. - Skip items with stale data and list them separately. - A setting to switch per-change alerts off when the digest is on. - Keep the "Analysis, not financial advice. Simulated trades only." line in the message. - Update the README.
Add a model comparison to the market watch app in this folder. Requirements: - Configure a second LLM in .env and run both on every scan, storing each read with its model name; only the primary model's read triggers alerts. - A page that shows, per symbol, where the two labels agree or differ, and the token cost of each. - A weekly count of agreement and of rejected reads per model. - Update the README.
What you lose
- Licensed live data across stocks, ETFs, futures, forex and crypto in one place
- The iPhone app and its phone alerts
- A public page with each of its own bots' closed demo trades, losses included
- The REST API, webhooks and MCP connection to AI tools such as Claude and ChatGPT
- Chat about any asset with live web checks
Why people still pay
People pay to have the assets they follow checked on a schedule, with the levels and the reason written out, without wiring data feeds and prompts themselves. It stays a tool for analysis: the product says it places no orders and that every decision is yours.
TradingWizard or your own copy?
Over
Your usage
12h ×$/h
TradingWizard, 1 yr$480
Your own copy, 1 yr$68running + build session
You keep$413pays for itself in the first month
Estimate · Crypto data is free and ntfy is free. The LLM costs cents per scan through an API, so ten assets on an hourly schedule add up to a few dollars a month, or nothing with a local model. The build is token usage on your agent plan.
Take your data, then cancel.
1 · Take your data out of TradingWizard
- 1In Settings, export your trade history as a CSV at any time.
- 2Export the whole account as JSON from Settings: profile, preferences, positions, trades, chats with its assistant and watchlists.
- 3The bots' rules and your per-asset settings aren't described as exportable: note the assets and timeframes you follow before you leave.
2 · Cancel TradingWizard
- 1On the web, open Settings, then Manage billing, and cancel; the FAQ describes it as one click with no contract.
- 2In the iOS app, cancel in your App Store subscription settings instead.
- 3During the 7-day trial, cancel before it ends and you pay nothing.
- 4You keep the plan until the end of the period you paid for, and it doesn't renew; yearly subscribers cancel the same way.
- Cancelling doesn't delete your data, according to the FAQ; the refund page says nothing about what happens to it afterwards, so email hi@tradingwizard.ai to ask.
- Fees are non-refundable except where the law requires, and partial months or years aren't refunded; App Store purchases follow Apple's refund process.
- Export the CSV and the JSON before the paid period ends, since bots and alerts stop with the plan.
The escape hatch
Don't feel like building it? These are already free.
TradingAgentsOpen sourceApache-2.0 multi-agent LLM framework for financial analysis; its authors present it for research and say it isn't financial advice.★ 112k · active Oct 2026 · one-click install
CLIVisit ↗AI Hedge FundOpen sourceMIT proof of concept of LLM agents analyzing stocks; its README says it is for educational use only and not for real trading.★ 64k · active Oct 2026 · one-click install
CLIVisit ↗FreqtradeOpen sourceGPL-3.0 crypto trading bot in Python with backtesting and a dry-run mode, controlled from Telegram or a web UI; rule-based strategies rather than an LLM.★ 56k · active Oct 2026 · docker to self-host
CLIVisit ↗
CLIVisit ↗AI Hedge FundOpen sourceMIT proof of concept of LLM agents analyzing stocks; its README says it is for educational use only and not for real trading.★ 64k · active Oct 2026 · one-click install
CLIVisit ↗FreqtradeOpen sourceGPL-3.0 crypto trading bot in Python with backtesting and a dry-run mode, controlled from Telegram or a web UI; rule-based strategies rather than an LLM.★ 56k · active Oct 2026 · docker to self-host
CLIVisit ↗
The numbers
TradingWizard pricing
| Plan | Monthly | Annual, per month | What you get |
|---|---|---|---|
| Pro | $40 | $29 | 5 demo bots (paused ones count), 10 alerts, unlimited chat with its assistant, 2,000 analyses a month and 5 live web checks a day, with a REST API and MCP connection listed on the pricing page. |
- Free tier
- Following a bot, setting an alert, asking the assistant or placing a demo trade all need a plan; the 7-day trial includes them.
- Billing
- One plan at $40 a month or $348 a year after a 7-day trial; cancel before the trial ends and nothing is charged. Fees are non-refundable except where the law requires, and subscriptions bought in the iOS app are refunded by Apple.
- Hidden costs
- The bots trade simulated money only and the product places no orders, so any real trading goes through a broker you pay separately. Plan limits cap bots at 5 and alerts at 10.
- Source
- tradingwizard.ai
Also on the list
Related apps
No free plan, only a 7-day trial that needs a card; $0 is charged today. You can browse without a plan.
Kinda. The core is a weekend build with the prompt on this page, but real gaps remain: Licensed live data across stocks, ETFs, futures, forex and crypto in one place, The iPhone app and its phone alerts.
About $40 a month on the Pro plan, $480 a year (checked 2026-10-19).
Licensed live data across stocks, ETFs, futures, forex and crypto in one place; The iPhone app and its phone alerts; A public page with each of its own bots' closed demo trades, losses included; The REST API, webhooks and MCP connection to AI tools such as Claude and ChatGPT; Chat about any asset with live web checks. If any of those matter to you, keep paying.
Yes: TradingAgents (open source), AI Hedge Fund (open source), Freqtrade (open source).