One API for stocks, crypto, perps, options, and prediction markets

Ask a developer to build an agent that works across stocks, crypto, options, and prediction markets, and much of the work is plumbing. Each venue has its own auth, symbols, sizing, fees, and settlement. A unified trading API reduces that integration glue to one agent-facing interface. Runtime status, account eligibility, and venue state determine what is currently available.

One symbology across five market types

With Felix, the agent learns one way to name an instrument and one way to size a trade. Sizes are always in US dollars, so there is no contract math, no lot sizes, no strike multipliers to reason about.

Crypto / perp     BTC, ETH, SOL          (or crypto:BTC)
Stock             NVDA, TSLA             (or stock:NVDA)
Option            option:BTC-20260711-70000-C
Prediction market pm:fed-cut-in-september

Inputs are forgiving. BTC, btc, and BTC-USD all resolve to the same thing, which matters when the caller is a language model that might phrase it any of a dozen ways.

The same request shape everywhere

Because the symbology is uniform, the supported integrations share one normalized order shape. Runtime, account, and venue state still determine whether a request can be accepted or filled:

POST /v1/orders {"instrument":"NVDA","side":"buy","size_usd":250}
POST /v1/orders {"instrument":"BTC","side":"buy","size_usd":50}
POST /v1/orders {"instrument":"option:BTC-20260711-70000-C","side":"buy","size_usd":40}
POST /v1/orders {"instrument":"pm:fed-cut-in-september","side":"yes","size_usd":20}

Positions, fills, PnL, and balances use one normalized read model and reconcile against venue truth. One normalized model spans five market types; every executed order is real money, and current availability comes from runtime status.

Why this matters more for agents than for humans

A human trader can tolerate several apps and habits. An agent pays an integration cost for every additional request shape and edge case. Uniformity reduces integration surface and inconsistent request shapes, but it does not guarantee model, system, data, or venue reliability. Fewer shapes let more of the workflow focus on the decision instead of plumbing.

Cross-market strategies become simpler to express

When market integrations share one normalized language, cross-market strategies become simpler to express. Hedge a crypto position with a perp. Express a macro view with a prediction market and a stock at once. Roll an option and rebalance spot in the same loop. Runtime and venue state still govern each action. See the account model in How to give an AI agent a trading account.