API guide
Financial data providers: how to choose one in 2026
Most buying guides are a wall of logos. The questions that decide whether a feed works for you are about licences, clocks and coverage, and you can answer all of them in a one-week trial.
On this page
- The five kinds of financial data providers
- Start from the job, not the vendor
- Licensing: personal, commercial, display and redistribution
- Coverage: test with your own symbol list
- Timestamps, units and evidence of quality
- Delivery, limits and the real price
- SLAs, support and the exit
- A scorecard you can copy
- Run a one-week trial before you sign
- Questions
Key takeaways
- Financial data providers range from exchange-direct feeds and institutional terminals to developer APIs, broker APIs and crypto aggregators.
- The licence often matters more than the price: check personal vs commercial use, display vs non-display use, and redistribution rights.
- Test coverage with your own symbol list, not with the provider's headline instrument count.
- Ask what every timestamp marks and whether bars are stamped at their start or their end.
- A one-week trial against an independent reference catches most problems before a contract does.
Financial data providers, also called financial data vendors or market data providers, collect prices, reference data and fundamentals from exchanges and other sources, and sell access to them as terminals, files or APIs. Choosing one comes down to five checks: the markets you need, freshness and timestamps, what the licence permits, how the data is delivered, and the cost at your real volume.
This is a vendor-neutral checklist, and it names no vendors on purpose: rankings go stale in a quarter, the questions do not. If you want a refresher on how a market data API is put together, start there. If you want named comparisons, our comparison pages set TickerLayer against specific alternatives.
The five kinds of financial data providers
Market data vendors sit at different distances from the market. The closer they sit, the faster and more official the data, and the heavier the licence and the engineering that come with it.
| Type | What you get | Freshness | Licensing burden | Typical buyer |
|---|---|---|---|---|
| Exchange-direct feeds | One venue's raw messages, full depth, official prices | Fastest | Heavy: exchange agreements, per-user and non-display fees | Trading firms, brokers, market makers |
| Institutional vendors and terminals | Consolidated global data, fundamentals, news, analytics | Fast | Heavy: enterprise contracts, per-seat pricing | Banks, funds, research desks |
| Retail developer APIs | REST and WebSocket access to prices and bars across assets | Real-time to delayed, varies by plan | Light to moderate: plan terms, sometimes exchange fees passed through | Startups, fintech apps, independent developers |
| Broker APIs | Prices bundled with a trading account | Tied to your account entitlements | Usually personal use only | Traders automating their own accounts |
| Crypto aggregators | Prices combined from many venues | Real-time | Light; venue terms vary | Crypto apps, wallets, analytics |
Free and scraped sources need their own warning. A library that scrapes a public web page feels like a free API until the page layout changes, the site blocks your server, or someone asks whether you had the right to use the numbers. Treat those as prototyping tools. The guide to free stock API tiers covers what the legitimate free options really include.
Start from the job, not the vendor
The most expensive mistake is buying for a use case you do not have. Write down four things before you open a single pricing page.
- Who sees the dataOnly you, your internal team, your paying users, or the public. Showing prices to other people is the biggest single driver of price and paperwork.
- What it drivesA chart, an alert, a backtest, a trading decision. Automated decisions call for stricter freshness and quality checks than a daily email does.
- Which instrumentsWrite the actual list: markets, symbols, asset classes, history depth. Headline counts are marketing; your list is the requirement.
- How often and how manySymbols streamed at once, requests per month, bursts at the open. This decides the plan, not the feature grid.
| If you are building | You probably need | Provider type |
|---|---|---|
| A personal dashboard or research notebook | Real-time or delayed prices and bars on a free or low tier | Retail developer API |
| A consumer app that shows prices to users | Display rights for every market you show, streaming for live screens | Developer API with commercial terms, or an institutional vendor |
| An algorithmic strategy | Low latency, clean timestamps, non-display rights | Exchange-direct or institutional |
| An AI agent or LLM tool | Simple REST, symbol discovery, market hours, explicit units | Developer API with an MCP server or pay-per-call access |
| Back-office valuation or reporting | Official closes, corporate actions, an audit trail | Institutional vendor |
If stocks are the core of your product, the stock API guide goes deeper on global listings, qualified symbols and bar semantics.
Licensing: personal, commercial, display and redistribution
Two providers can send you identical numbers under very different rights. Read the licence for four distinctions before you read the price.
- Personal vs commercialPersonal or research plans usually forbid using the data inside a product, even an internal one. Commercial terms cost more because they allow it.
- Display vs non-displayDisplay means a person sees the number. Non-display means software consumes it, for example to trade. Exchanges often license and price the two separately.
- RedistributionShowing prices to your users, passing them through your own API, or putting them in a download is redistribution, and it needs explicit permission.
- Derived dataAn index, a signal or a chart built from licensed prices can still fall under the original licence. Ask how derived works are treated.
Our own plans show the pattern: Individual plans are for personal and research use only, and Business plans include commercial use. Whoever you pick, get the answer to "may my users see this?" in writing before you build the screen.
Also ask what the data is. TickerLayer prices are derived and indicative, built by aggregation rather than taken from an exchange's official feed, and our market data disclaimer says so plainly. That suits apps, dashboards, research and agents. If a regulation or a contract requires official exchange prices, you need a provider licensed to supply exactly that.
Stock market data providers and exchange fees
Equities add one more layer. Exchanges charge for their real-time data, and many classify every end user as professional or non-professional, with very different monthly fees for each. Stock market data providers that pass exchange data through will ask you to declare that status for each user who sees live prices, and the paperwork grows with your user count. Delayed data usually carries lighter terms, which is one reason so many consumer apps show stock prices 15 minutes late.
Coverage: test with your own symbol list
Headline numbers, our own "150K+ symbols" included, say little about whether your 40 instruments are covered in the right markets with the history you need. Paste your real list into each provider's coverage tool or run it through their API. Ours is the symbol coverage checker: it normalizes formats (KO to US:KO, EUR/USD to EURUSD), flags ambiguous tickers and marks anything not covered.
Coverage questions for your list
- Every symbol resolves to the listing you mean, in the market you mean.
- History goes back far enough at the intervals you need, for example a year of 1-minute bars.
- Pre-market and after-hours data are included if your users trade them.
- New listings appear quickly, and delisted symbols keep their history.
- Splits and other corporate actions are handled the way your charts expect.
- Market hours and holidays are available by API, not only on a web page.
Timestamps, units and evidence of quality
Timestamps are where careful providers and careless ones look identical until the day you compare them. Ask what each time field marks: the market event, the moment the provider received it, or the moment it was sent to you. Ask whether bars are stamped at their start or at their end, since a 1-minute bar stamped at its end sits 60 seconds later than the same bar stamped at its start. Ask which date a daily bar carries, and in which time zone.
| Question | A good answer looks like |
|---|---|
| What does the quote timestamp mark? | A named event, in UTC, with its precision stated |
| Are bars stamped at the start or the end? | One documented rule for every interval, for example "t is the bar start" |
| What unit is each price in? | A unit per symbol, including exceptions such as sugar in US cents per pound |
| How are bad prints handled? | A described filter, and a way for you to report a bad value |
| How is quality measured? | A published method against an independent reference, with dates and sample sizes |
Quality claims deserve the same skepticism as coverage claims. Ask for a method, not adjectives: deviation from an independent reference in basis points, how fresh prices are, whether bars rebuild identically on a second request. We publish ours on the data quality pages as observed behaviour under a stated method, not as a guarantee. To test any provider yourself, the bad ticks guide shows how to catch stale and spiking prices in code.
Delivery, limits and the real price
Delivery and limits decide the real bill. A REST-only plan that looks cheap can cost more than a streaming plan once you poll 50 symbols every few seconds. Do the arithmetic with your own numbers before you compare list prices:
requests per month = symbols × (seconds of coverage per day ÷ interval) × days
- symbols
- Instruments you poll.
- interval
- Seconds between two polls of the same symbol.
- seconds of coverage
- 23,400 for a 6.5-hour US session; 86,400 for markets that trade around the clock.
- days
- About 21 trading days a month, or 30 for crypto.
Then price the whole cart: per feed, per market, per displayed user, overage and history depth. TickerLayer sells plans per feed (crypto, forex, stocks by market, and so on), from $49 a month on Individual and $449 on Business, with 3,000 free REST requests a month to evaluate. Other providers price by seat, by exchange or by request. None of these models is wrong; each rewards a different usage shape, so model yours. The full grid is on our pricing page.
SLAs, support and the exit
An SLA usually covers the availability of the service, not the correctness of each price. Read what it measures, how it is measured and what you receive when it is missed. Then ask about the parts of the relationship that matter on a bad day:
- Support: who answers, how fast, and whether they can see your requests when you report a bad price.
- Change notice: how long before a field, symbol or endpoint changes, and how you hear about it.
- Contract: term length, auto-renewal, and whether you can downgrade without penalty.
- Exit: whether you may keep the historical data you downloaded after you leave.
A scorecard you can copy
Score two or three finalists on the same sheet. The weights below suit a product that shows prices to its users; shift weight toward latency for trading and toward history depth for research.
| Criterion | Weight | What earns a 5 |
|---|---|---|
| Coverage of your list | 20 | Every symbol resolves, with the history you need |
| Licence fit | 20 | Your exact use is permitted, in writing |
| Timestamp and unit clarity | 15 | Every field documented; bars and dates unambiguous |
| Measured quality | 15 | Published method, and your own week of checks agrees |
| Delivery and limits | 10 | REST and streaming, with limits visible in response headers |
| Total cost at your volume | 10 | Fits the budget at twice your forecast |
| Developer experience | 5 | Clear errors, runnable examples, symbol discovery by API |
| Support and exit | 5 | A named contact, change notices, data you can keep |
Run a one-week trial before you sign
- Load your real listResolve every symbol through the provider's symbol endpoints. Anything ambiguous or missing shows up on day one.
- Cross a session boundaryRecord an open, a close and a night. Timestamp, snapshot and holiday bugs cluster at the edges of a session.
- Compare against a referenceSample prices against an independent source and measure the gap in basis points. One bad hour tells you more than a week of good ones.
- Break it on purposeSend a bad symbol, exceed the rate limit, drop the connection. Read the errors you get back and judge whether your code could act on them.
- Price the real monthMultiply your measured usage by the price list, then double it for growth.
Green flags
- Timestamps, units and bar rules documented field by field.
- Rate limits visible in response headers.
- A quality method published with dates and sample sizes.
- A licence that names your use case explicitly.
- Symbol lists available by API.
Red flags
- "Real-time" with no definition and no per-market exceptions.
- Coverage counts you cannot check against your list.
- Licence terms shown only after you sign.
- No way to report a bad price.
- Errors that return 200 with an empty body.
Questions
What are financial data providers?
Companies that collect prices, reference data and fundamentals from exchanges and other sources and sell access to them through terminals, data files or APIs.
How much does market data cost?
It ranges from free tiers with a few thousand requests a month, through developer plans priced per asset class, to enterprise contracts priced per user and per exchange. TickerLayer plans start at $49 a month per feed.
What is the difference between display and non-display data?
Display data is shown to a person on a screen. Non-display data is consumed by software, for example an algorithm that trades on it. Exchanges often license and price the two separately.
Can I use free financial data in a commercial app?
Only if its licence says so. Many free tiers, and every scraped source, restrict commercial use or redistribution, so read the terms before you ship.
How do I evaluate a market data provider?
Run a one-week trial with your own symbol list: resolve every symbol, record a market open and close, compare prices with an independent reference, and trigger the error paths on purpose.