Getting started
Guides
Nine things developers build on the computed table, each reduced to a handful of calls. Every recipe is runnable: the feed, the warehouse, and the indicator math are already done.
The recipes
Simple to complex. Each guide is a few curls and a code block — the same grammar throughout.
| Guide | What it shows |
|---|---|
| Build a stock screener | One SQL WHERE clause screens every ticker server-side: live, as of the past, or as a push. |
| Build a custom indicator | Define an indicator once as a SQL predicate and it works everywhere: screens, alerts, webhooks, series, and history. |
| Build a backtester | As-of scans give point-in-time candidate sets with no look-ahead or survivorship bias; aligned series price the trades. |
| Build an ML dataset | Point-in-time features from ?asof=, labels from aligned series, delisted tickers included: leakage-free training data by construction. |
| Build a stock alert bot | Write the condition in SQL, get a push when it fires: webhook to your bot or straight to Discord with no server at all. |
| Build a trading algorithm | Define an entry condition in SQL, backtest it as of any past moment, run it live as a webhook: entries and exits from one delivery. |
| Build an AI trading agent | The market as native MCP tool calls: computed answers, not raw feeds, so the model spends context on decisions. |
| Build a portfolio tracker | Holdings as a universe: price every position in one call, chart from aligned history, and get alerts scoped to what you own. |
| Build a market dashboard | One row per ticker with every signal, aligned series for charts, websocket push for freshness: three calls. |