How to stream real-time stock quotes

The short way

Alpaca Market Data streams live US trades on a free key, over WebSocket, with no card:

from alpaca.data.live import StockDataStream

stream = StockDataStream("YOUR_KEY_ID", "YOUR_SECRET")

async def on_trade(trade):
    print(trade.symbol, trade.price, trade.size, trade.timestamp)

stream.subscribe_trades(on_trade, "AAPL", "MSFT")
stream.run()

That is genuinely live. It is also genuinely partial: the free Basic plan carries the IEX feed and nothing else, caps the stock WebSocket at 30 symbols, limits you to 200 requests a minute and withholds SIP data from the last 15 minutes. Finnhub is the same shape from the other direction — the free key streams live US trades for up to 50 symbols, and returns no OHLC candles at all.

Both are enough to build the thing and see it move. Neither is the consolidated tape.

What the options are

Free, live, narrow. Alpaca Market Data and Finnhub as above. Different limits, same bargain: a real feed with the coverage or the symbol count cut down to where it costs the vendor nothing much.

One flat price for everything US. Alpaca's Algo Trader Plus is $99 a month for the full SIP and OPRA at 10,000 requests a minute with unlimited stock subscriptions, with no per-asset-class pricing. That flatness is the product.

A direct feed with a ladder. Massive, formerly Polygon.io, sells US equities, options and futures from a direct exchange feed. Read the ladder carefully: the free Basic plan is end-of-day at five calls a minute with no WebSocket at all, $29 Starter buys the WebSocket but is still 15-minute delayed, and real-time trades and quotes do not start until the $199 Advanced plan.

Global rather than US. Twelve Data covers 85 markets, and its streaming has the same shape of catch: the free and $29 plans carry trial WebSocket credits limited to one connection, eight symbols and the vendor's own trial symbol list. Streaming a symbol you picked yourself starts on Pro at $99, or Venture at $149 if you are a company.

Into a socket on your own machine. IQFeed is not a hosted API at all — it is a Windows client you launch, which you then talk to over TCP sockets on localhost, one port for Level 1 streaming and another for market depth. $108.15 a month plus a one-time $50 startup fee, and every real-time exchange's own fee on top of that, itemised. Any language that can open a socket works; there is no HTTP endpoint.

The venue feed, unaveraged. Databento is where you go when the normalisation is the thing you want to control, and it itemises the exchange fees rather than averaging them into a plan.

Where this breaks

You are not really buying from the vendor. Nearly all of what a live quote costs is set one layer above, by the exchange, and in US equities the price list is a public document. The NYSE pricing guide dated 14 May 2026 prices the NYSE Integrated feed at $16.00 a month per non-professional device and $78.00 per professional one — the same bytes, five times the price, decided by a question about you rather than about the data.

Professional versus non-professional. The test is the exchange's and lives in its data policy, not in your vendor's signup form. Broadly: a natural person, using the data for personal non-business purposes, not registered or qualified with a securities regulator, not using it for any business. A company of one does not qualify. A screen at work does not qualify. If any part of that is shaky, you are professional, and you are on the other column of every schedule.

Display versus non-display. If a program acts on the quote and no person reads it, that is non-display use, licensed separately and usually far higher, because it is billed per firm or per platform rather than per screen. On the same NYSE guide, NYSE Integrated non-display is $22,400 a month in each of three categories. Nobody looking at the data is the expensive case, not the cheap one. Twelve Data's free tier states this the other way round — it is licensed internal non-display, meaning evaluation and development, not production and not shown to users.

Redistribution ends the side project. The moment a live quote reaches anybody who is not you — a public page, a bot, a free tier for friends — it is redistribution, and it is its own line: $4,400 a month for NYSE Integrated on the May 2026 guide, beside an $8,400 access fee.

The per-asset-class surcharges. IQFeed is the clearest published example of how this stacks, because it prints every layer: $108.15 for the core service, $60.90 for real-time equity options as a surcharge, and then OPRA's own $7.35 non-professional fee on top of that. Most vendors bury the same structure in one number, which is fine until you change asset class.

Why real-time stock data is so expensive is the full version of all five of these, with the schedules.

If you outgrow this

Ask whether delayed breaks anything. Screeners, trackers, dashboards and every backtest are fine on data that carries no per-user fee at all, and that is most of what gets built. The 15-minute line is a licensing decision, not a technical one.

Check what your broker already entitles you to. Brokerages buy non-professional entitlements in bulk and pass them through for a few dollars or nothing. Look there before buying the same feed twice.

Buy the tape rather than the direct feeds if latency is not your product. The consolidated tape is the regulated, cross-venue view; it is slower and dramatically cheaper per user than the proprietary feeds, and a vendor bundling it is buying it at plan rates for you.

Count symbols, not requests. When a streaming plan starts to bind, the field to compare is the simultaneous-subscription cap — 30 on Alpaca free, 50 on Finnhub free, 250 on Finnhub Basic, unlimited on the paid tiers of both.

Real-time market data tools is the catalogue narrowed to the cards serving a live feed, and market data APIs is the wider listing.