Looking for the finished bot instead? Take the Tradingale Runner. This recipe is for building your own tool on the same data.
Your own companion, in one evening
Tradingale is a data layer. We compute the Martingale Score and the Startingale across 250+ crypto pairs and 100+ US stocks, and we serve them two ways: a REST API and an MCP server. What you build on top of that data is yours. Your machine, your code, your decisions.
This recipe shows how fast “yours” can happen. With Claude and one API token, a personal dashboard that reads Tradingale data goes from idea to running on your laptop in a single session.
What the data is, and what it is not
Every instrument in the catalog carries the same descriptive fields:
| Field | What it describes |
|---|---|
| martingale_score | Statistical compatibility of the instrument with the martingale model, 0 to 5 |
| startingale | Alignment of current conditions with the model’s entry structure, 0 to 5 |
| score_price_behavior, score_liquidity, score_market_activity | The three components behind the score |
| delta_price | The price spacing the model associates with the instrument, as a fraction |
| nb_rounds | The depth of the model’s structure, 4 or 5 rounds |
| multipliers | Round-to-round size ratios of the model, currently between 2 and 2.5 |
| initial_bet_ratio | The fraction of a sequence budget the model assigns to round 1 |
| active_on_binance, active_on_kraken_pro | Where the instrument trades |
| timestamp | When the data was computed |
Notice what is absent: there are no quantities and no amounts anywhere. Every field is a score, a count, or a dimensionless ratio. The data describes how the model reads each instrument. It does not tell you, or your software, what to do. That boundary is the product.
Two ways to plug in
MCP, for agents
If you use Claude, one command connects it:
claude mcp add --transport http tradingale https://tradingale.com/mcp
Seven read-only tools become available: instrument lookup, search and filters, top lists, and market overviews for crypto and stocks. The full walkthrough is in the Connect Claude in 60 seconds recipe.
REST, for code
Same data, plain JSON:
curl -H "Authorization: Bearer $TRADINGALE_TOKEN" \ "https://tradingale.com/api/martingale/crypto?martingale>=4&limit=20"
Both run on the same token and the same monthly quota. Tokens live in Settings > API.
The prompt
Paste this into Claude Code, ChatGPT, Gemini, Cursor or any coding agent you like, point it at your token, and let it work:
Build me a small local web dashboard (single Node.js server, no database) that uses the Tradingale REST API with my bearer token from the TRADINGALE_TOKEN env var. Endpoints: - GET https://tradingale.com/api/martingale/crypto?limit=200 (list, filterable with martingale>=X, startingale>=X, exchange=binance|kraken) - GET https://tradingale.com/api/martingale/stocks?limit=200 Each instrument has: martingale_score, startingale, score_price_behavior, score_liquidity, score_market_activity, delta_price (fraction), nb_rounds (4 or 5), multipliers (array of round-to-round ratios), initial_bet_ratio, timestamp. The dashboard should: 1. Show crypto and stocks tabs, sortable by martingale_score and startingale. 2. Let ME define watch thresholds per instrument (my numbers, stored in a local JSON file) and highlight rows when MY thresholds are crossed. Refresh every 10 minutes and respect the API quota. 3. When I click an instrument, show its model structure: delta_price spacing, nb_rounds, multipliers, initial_bet_ratio, and a sequence plan preview computed from a budget number I type into a field (all arithmetic client-side). 4. Show the data timestamp on every view so I always know how fresh it is.
This prompt produces a dashboard that watches data and renders arithmetic. Want it to do more? Extend it however you like, including toward your own exchange account: that extension is your software, running on your keys, under your sole responsibility. Our part stops at the data.
Why this shape
We publish how the model behaves when applied mechanically, with the full history on our track record page (simulated execution). We publish the descriptive metrics it produces, every day, on every instrument in the catalog. And we stop exactly there. The reading is ours; the decision is yours. That is what “data layer” means, and it is the only version of this product we want to build.
Tradingale never places orders, holds funds, or gives advice. What you build and run is yours.
Data for informational purposes only. Not investment advice.