SDKs

The SDKs sit next to the REST API and the MCP server: same data, same plans, same keys. They handle retries and rate limits, page through your plan's history limits, and verify webhooks, so you write the analysis, not the plumbing.

Install

Python 3.9+

pip install https://docs.madjik.io/sdk/madjik-0.2.0-py3-none-any.whl
pip install "madjik[pandas] @ https://docs.madjik.io/sdk/madjik-0.2.0-py3-none-any.whl"    # with DataFrame output

TypeScript / JavaScript (Node 18+, zero dependencies)

npm install https://docs.madjik.io/sdk/madjik-0.2.0.tgz

Google Sheets

Download madjik-sheets-Code.gs, paste it into Extensions → Apps Script in your sheet, reload, then Madjik → Set API key.

These files are served from docs.madjik.io (MIT licence) and will also be listed on PyPI and npm, after which the commands shorten to pip install madjik and npm install madjik. Verify what you download against the SHA-256 values below.

Python

from madjik import Madjik
mj = Madjik()                                    # reads MADJIK_API_KEY
mj.metrics(as_dataframe=True)                    # every metric in ONE call
mj.definition("M10004")                          # inputs, lookback, version (no key needed)

# history at scale: a dry run first, then a cached, quota-aware panel
mj.plan_calls(["M10004", "M10005"], start="2026-09-01", interval="1h")
df = mj.panel(["M10004", "M10005"], start="2026-09-01", interval="1h", live_only=True, lag=1)

panel() pages through your plan's points-per-call limit, paces calls to the speed limit, caches finished history chunks on disk (an interrupted or quota-limited run resumes the next day), and returns one UTC-indexed table. live_only=True gives point-in-time history and lag=1 shifts values so a value is only used after its timestamp. join_asof() attaches prices without look-ahead. A 429 waits for Retry-After and retries.

TypeScript / JavaScript

import { Madjik } from "madjik";
const mj = new Madjik();
const all = await mj.metrics();                                                        // every metric, one call
const ts = await mj.timeseries("M10004", { start: "2026-09-01T00:00:00Z", liveOnly: true });

Typed, Node 18+, the same retry behaviour, and an Express handler for webhooks.

Receiving webhooks (Pro and Enterprise)

# Python (FastAPI)
from madjik.webhooks import fastapi_handler
app.post("/madjik")(fastapi_handler(SECRET, lambda ev: print(ev.metric_id, ev.value)))

// Node (Express)
import { expressHandler } from "madjik";
app.post("/madjik", express.raw({ type: "application/json" }), expressHandler(SECRET, async (ev) => { /* ... */ }));

Both verify X-Madjik-Signature-V2 with a 5-minute replay window and deduplicate on X-Madjik-Event-Id. See Webhooks for the delivery contract; mj.test_webhook(id) sends a sample event and re-enables a webhook that was disabled after repeated failures.

Google Sheets

=MADJIK("M10004")                          latest value (or "signal", "method", "confidence", "timestamp", ...)
=MADJIK_SERIES("M10004", 30, "1d", TRUE)   last 30 days, live-only (point-in-time safe)
=MADJIK_FRESH("M10004")                    when it was last refreshed (no key needed)
=MADJIK_STATUS()                           127/129 metrics fresh (no key needed)

All metrics are fetched in one call and cached for a time that follows your plan (Free about 3 h, Starter about 17 min, Pro about 2 min), so a large sheet costs one call per refresh.

Downloads

Madjik metrics are informational indicators, not investment advice.