One ID for every real-world asset, on every venue.

The problem

Same market, different names

Real-world assets have tickers, ISINs and exchanges of record. Their perpetuals do not. Every venue invents its own scheme: one names TSLA by ticker, another by a numeric market index, a third behind a deployer's dex: prefix. The same ticker can also mean a stock on one venue and an unrelated token on the next.

Join them by hand and the code works until it silently doesn't. Atlas does the mapping once, in the open, for everyone: each contract gets an Atlas ID, and each asset an Atlas asset ID that is the same on every venue.

The symbology

Two IDs: the contract and the asset

The Atlas ID (internal_id) names the contract as the venue lists it. It is lossless, so it maps back to the venue's own instrument.

<type>-<symbol>-<denominator>[:<margin>][-<yyyymmdd>]

The unified ID writes the same contract in Atlas asset IDs, so it reads the same on every venue. Hyperliquid's xyz:TSLA and Bybit's TSLAUSDT both trade the asset TSLA.

type

spot, perpetual or future.

symbol

The base as the venue lists it. 1000PEPE stays 1000PEPE, and a HIP-3 dex keeps its prefix.

denominator

The quote or settlement currency the price is in.

:margin

The collateral asset. Derivatives only; it tells linear from inverse.

-yyyymmdd

The delivery date. Dated futures only.

underlying

equity, commodity, forex, index, pre_market or crypto, from the venue's own metadata.

Each record also holds the contract size, margin asset, asset class, the Atlas asset IDs of every leg and the listing window: first_capture and, once it stops trading, end_date.

Quick start

Local lookups, no key

Python 3.12 or later. The snapshots ship with the package, so every lookup is a local dictionary read with no network call.

pip install git+https://github.com/aperiodic-io/atlas

Prefer plain data? Every snapshot is published as JSON next to this page, for example data/avantis-perps.json, and the whole catalogue as catalog/instruments.json.