Simply Wall St GraphQL API: What Returns Fundamentals Instead

Copy this line to your agent to pull a company's income statement.

set up https://monid.ai/SKILL.md and use stockanalysis /get_stock_financials_income for a ticker

People search for the Simply Wall St GraphQL endpoint because they found it in browser devtools and want to know whether they can build on it. The short answer is no: it is an internal endpoint behind a consumer product, undocumented, unversioned and not offered for third-party use. The useful answer is what does return the same fundamentals, so this guide measures one on 2026-09-02 and shows the two things about its response that will bite you, running through Monid, the OpenRouter for agent tools.

Does Simply Wall St have a public API?

Not one you can build on, and the distinction between "exists" and "usable" is the whole question.

What people find

Open the network tab on a Simply Wall St company page and you will see GraphQL requests returning the data behind the snowflake chart. It looks like an API because technically it is one. It is the private interface between their web client and their backend.

Why that is not a foundation

An internal endpoint carries no compatibility promise. The schema changes when their frontend changes, there is no deprecation notice because there is no external contract, and access controls can appear overnight. Building a product on it means your product breaks on their release schedule, silently, and you find out from a customer.

There is also the terms question, which is the part people skip. Consumer analytics products license their underlying data from market data vendors, and those licences restrict redistribution. Pulling their GraphQL endpoint is not merely fragile, it is redistributing someone else's licensed data.

What to do instead

Use a source whose business is supplying the data, and check that it tells you where its numbers came from. That last part is unusual enough to be worth a whole section below.

📖 See also Crunchbase API Alternatives for Funding Data

Why is revenue0 not last year's revenue?

Because the response is column-oriented and the first column is TTM, and this is the single easiest way to produce a wrong number from correct data.

The shape

stockanalysis/get_stock_financials_income for NVDA on 2026-09-02 returned a financialData object holding 26 line items, each an array running parallel to a shared datekey array:

{
  "datekey":   ["TTM", "2026-01-25", "2025-01-26", "2024-01-28", "2023-01-29", "2022-01-30"],
  "fiscalYear":["2027", "2026",       "2025",       "2024",       "2023",       "2022"],
  "revenue":   [302970000000, 215938000000, 130497000000, 60922000000, 26974000000, 26914000000]
}

The line items are datekey, fiscalYear, fiscalQuarter, revenue, cor, gp, sgna, rnd, goodwillIntangibleAmortization, opex, opinc, interestExpense, interestIncome, currencyGains, otherNonOperating, ebtExcl, gainInvestments, mergerRestructureCharges, otherUnusualItems, pretax, taxexp, netinc, netinccmn, sharesBasic, sharesDiluted and epsBasic.

The trap

revenue[0] is 302,970,000,000 and that is trailing twelve months, not fiscal 2026. Fiscal 2026 is revenue[1], 215,938,000,000. Both are correct numbers. Reading index 0 as "the latest annual figure" mixes a trailing figure into an annual series, and because TTM is always the largest number in a growing company, the error looks like growth rather than like a bug.

The fix

Never index positionally. Find the period you want by its label:

i = data["datekey"].index("2026-01-25")   # or filter out "TTM" first
revenue_fy26 = data["revenue"][i]

Or drop TTM up front when you want a clean annual series:

keep = [i for i, d in enumerate(data["datekey"]) if d != "TTM"]

Why this shape at all

Column-oriented responses exist because that is how a financial statement renders: a table with line items down the side and periods across the top, so the API returns the table rather than a list of records. It is efficient and it matches the source document. It is also the opposite of what most JSON consumers expect, which is an array of period objects each carrying its own fields.

Converting once, at the boundary, is usually worth it:

rows = [
    {k: v[i] for k, v in data.items() if isinstance(v, list)}
    for i in range(len(data["datekey"]))
]

After that every downstream filter, sort and join works the way the rest of your code does, and the TTM row is a row you can drop rather than an index you must remember.

The other thing to notice

The response carried full_count: 10 while returning six periods. Ten periods exist; you got six. If you are building a long-run series, that field is telling you to page rather than assuming you received everything, and it is the sort of quiet ceiling that produces charts starting in the wrong year.

How do you pull fundamentals for a ticker?

Three steps. Discovery costs nothing.

For agents

Grab an API key at app.monid.ai, then paste this to your agent and hand it the key:

set up https://monid.ai/SKILL.md

It learns the whole discover, inspect, run workflow itself. More in the agent quickstart.

For humans

npm install -g @monid-ai/cli
monid keys add -k <your-api-key> -l main

Step 1. Pull the income statement

What it does. Returns the full statement across periods.

The endpoints. stockanalysis/get_stock_financials_income, billed per call.

The call.

monid run -p stockanalysis -e /get_stock_financials_income --query '{"symbol": "NVDA"}'

The parameter is symbol. Passing ticker returns a 422 saying Parameter 'symbol' is required, which is the good kind of failure: it names the field and stops. Worth appreciating, because the same category also produces responses that return a full field list with nothing in it, which is what the Amazon ASIN guide documents.

What comes back. The 26 line items above, six periods, source: "spg".

What it costs. A cent per call. Current figures at monid.ai/tools.

Step 2. Pull the statistics for ratios and position

What it does. Gives you the derived measures rather than the raw statement.

The endpoints. stockanalysis/get_stock_statistics, per call.

The call.

monid run -p stockanalysis -e /get_stock_statistics --query '{"symbol": "NVDA"}'

What comes back. Fourteen sections: trust, valuation, dates, shares, ratios, evRatios, financialPosition, financialEfficiency, taxes, stockPrice, shortSelling, incomeStatement, balanceSheet and cashFlow. That covers most of what a screener needs without you computing it from the statements.

What it costs. Same order.

Step 3. Read the provenance before you use the number

What it does. Tells you whether the figure is current, which no amount of schema validation will.

The call. Already in the response, in trust. See the next section, because it is the most useful thing here and almost nobody looks at it.

📖 See also Company News API: Press Releases Without the Newswire Contract

How do you know whether a financial number is current?

You read the metadata, and the reason this deserves a section is that most financial endpoints do not give you any.

What came back

The statistics response carried a trust object:

{
  "sources": [{ "name": "S&P Global Market Intelligence", "short": "S&P Global" }],
  "freshnessLag": "hours",
  "updateFrequency": "daily",
  "lastUpdated": 1788359436257,
  "lastChecked": 1788359436257
}

That timestamp is 2026-09-02T14:30:36Z, roughly when the call was made.

Why this matters more than the numbers

A financial figure without a date and a source is not data, it is a rumour with decimal places. Two failure modes disappear when provenance travels with the value. You can tell a stale cache from a company that has not reported, because lastUpdated distinguishes them. And you can tell whether two numbers are comparable, because they name the same upstream or they do not.

What to do with it

Store lastUpdated next to every figure you persist, and alert when it stops moving. A source that silently freezes looks exactly like a quiet market, and the timestamp is the only thing that separates them. This is the same discipline as watching a block rate rather than an error rate, argued in the proxy guide.

What most sources give you instead

Nothing, usually. The common shape is a bare number with no indication of when it was computed or which vendor supplied it, which means the only way to detect a frozen feed is to notice that a figure you expected to move has not. That detection happens weeks late and usually via a person, not an alert.

The practical consequence is a schema decision. If your source supplies provenance, carry it through your own storage layer rather than discarding it at the parse step, because the moment you drop it you have converted a checkable number into an unfalsifiable one. We made this argument about pipelines generally in buy versus build for firmographics: the metadata is often worth more than the marginal field.

The honest caveat

freshnessLag: "hours" and updateFrequency: "daily" are the provider's own claims about themselves. They are useful and they are not audited. If a number carries money risk, verify a sample against the company's own filing rather than trusting the label.

Which endpoint should I use for which job?

Every row was verified with monid inspect on 2026-09-02. The table gives billing shape rather than figures; shape drives design and current numbers live on monid.ai/tools.

The filings row is the one to reach for when a number looks wrong. A filing is the primary source, and checking one figure against it costs a call and settles the argument, which is cheaper than either trusting or distrusting an aggregate.

When do you need a licensed market data feed?

Three cases.

You are trading on it. Execution needs exchange-licensed real-time data with a latency guarantee and an audit trail. Fundamentals endpoints are for research, not for a trading loop, and the freshness label above is a claim rather than an SLA.

You are publishing figures to customers. Redistribution is exactly what market data licences govern. If your product displays financials to paying users, the licence is a product requirement.

You need point-in-time history. Financial statements get restated, and most sources show you today's version of the past. Backtesting against restated history quietly overstates results because you are using numbers nobody had at the time. Point-in-time databases exist for this and cost accordingly.

And the disclosure: this is Monid's blog and we sell per-call access to the endpoints above. The three cases here are cases where you should buy something else, and the caveat about freshnessLag being self-reported applies to data we are selling you.

Conclusion

There is no public Simply Wall St GraphQL API. The endpoint in your devtools is their web client talking to their backend, and building on it means shipping a product that breaks on someone else's release schedule while redistributing data they licensed from a third party.

What does work returned, on 2026-09-02, a 26-line income statement across six periods and fourteen sections of derived statistics, with the source named and a timestamp attached. Two things about that response decide whether you get right answers from it. The first column is TTM rather than the latest fiscal year, so index by label and never by position. And full_count reported ten periods while returning six, so page rather than assume.

Free next step: run monid discover -q "stock fundamentals financials" and read the field lists. Discovery is free, and knowing that a response is column-oriented before you write the parser saves the exact off-by-one this article is about. Start at monid.ai.

FAQ

Last updated September 2026.