5 / 5 steps done · 5.6s

A visual canvas for the next generation of traders.

Fynlab puts charts, fundamentals, news and SEC filings on one canvas. Ask a question and a board comes back, every figure with its source. All your analysis in one place.

In alpha · no credits, ever · built on open source
Earnings rundown · CSCO5 nodes
One board

Price and the filings, side by side.

Every figure is a card and the chart is one more of them. Put it next to the financials and the price, and all three read the same company without being told twice.

NVDA · your canvas4 nodes
You start from a preset

A preset arrives built. From the palette you add whatever it is missing, and the canvas keeps what you put on it, ready for the next company.

The chart is interactive

Change the interval, draw on it and add indicators inside the card itself, without leaving the board.

One ticker for the whole board

Wire a node in and it inherits the canvas's ticker. You type it once, at the top.

01 · it never starts empty

A preset is a process, not a subject.

Every canvas begins from a runbook and a company, and arrives wired and run. Change the company on the first node and every step below it runs again — same canvas, same process, different subject.

02 · the nodes run

Each step has its own state and its own clock.

Fetch, compute or write — with a timing, an output and a source. Re-run one without redoing the canvas. Nothing refreshes on its own, because each step is a request to a filing.

03 · numbers it cannot invent

The gap is reported, never filled.

Figures are fetched and shaped by fixed code, then labelled with the filing they came from. When a value genuinely is not published, the node says unavailable and why — it never substitutes another number.

Who it is for

For the next generation of traders.

The ones who read the chart and the filings in the same sitting, use AI to move faster, and still want the document behind every number.

The chart and the filings, together
Price used to live in one tab and the filings in another. Here they sit on the same board, reading the same company, and you switch company from the card at the top.
You check what the AI tells you
The AI saves you the fetching. Every sentence it writes carries the node it came from and every figure the filing it came from, so checking it costs one click.
No terminal, no institutional seat
These tools were built for trading desks. This one you open yourself, keep your own process, and reuse it on the next company.
Your positions

Every holding on the same board.

A canvas is a board: every figure is a card and the arrows say what feeds what. Put the tickers you follow on one of them, each with its own filings, financials and price, and ask about all of them at once.

Your holdings · NVDA, AMAT, CSCO5 nodes
One branch per holding

Each ticker brings its own cards and runs on its own. If one company fails to fetch, the others carry on and it names the one that did not.

Questions scoped by you

Ask reads the cards you wire into it and no others. Attach all three holdings and it answers across three; attach two and it compares those two.

The charts on one screen

The advanced chart draws several tickers at once, so you compare them by eye without opening another tab. Any holding you have not refreshed says so on its own card.

Build & ask

Describe the analysis. Then question it.

Build turns a sentence into wired nodes, already pointed at the company on your canvas. Ask answers across the steps you connect to it — so a question about margins reads the margin nodes and nothing else.

Build proposes the nodes and the wires between them. You get a working subgraph, not a suggestion to go read the docs.
Ask is scoped by the edges you draw — its attachments are its incoming edges. You pick what a question reads by wiring it, never by describing context.
Figures are never the model's to produce — see the method for where each one comes from.
Presets

Start from a process, not a blank page.

A new canvas opens with three options: a question, a company, or a blank board. The presets are the company option: pick one, type a ticker, and the canvas opens built and already run.

01Earnings rundown
A company just reported: the quarter against prior periods, trailing margins, and the filing and price context around the report.
7 nodesEDGARNasdaq
02Due diligence
Three years of filings and numbers, all three statements, margins, returns, multiples and what the insiders filed — and what this run cannot tell you, named.
13 nodesEDGARNasdaq
03Weekly review
Per holding: what was filed, how the numbers stand and how the price moved — split into what changed and what is routine.
5 nodesEDGARNasdaq
04Dividend and buyback record
What a company pays out and buys back, the cash flow and debt beside it, and the ratios it will not compute when two windows do not line up.
7 nodesEDGAR
05Filing diffnext
This year's annual report against last year's: new risk factors, changed language, segment shifts.
06Thesis stress testnext
Paste your written thesis; the canvas argues against it using the company's filings and numbers.
07Blank canvas
Here if you want it — just never the default. Wire your own nodes from the palette.
Catalog · 26 nodes

Everything you can put on a board.

Presets are the starting point. From the palette you add any of these, and what you build stays built.

Technical7
Simple moving averageExponential moving averageRolling VWAPMACDStochastic oscillatorRelative strength indexAverage directional index
Fundamental11
Company profileCredit profileDividend informationBuyback trackerAnnouncementsProfitability ratiosBalance sheetValuation multiplesGrowthCash flowReturns on capital
Market data6
Price performanceCorrelationSector contextSector mapShort interestAdvanced chart
Ownership and insiders1
Insider flow
Ask1
Ask

Every node names its source and links to it. When a company has not published a figure, the node says so and why, and no other number appears in its place.

Status

Where this actually is.

If you are going to put your name next to it, you deserve to know what is built and what is not. This section is the short answer.

Deployed and working, and genuinely in alpha. Nobody outside the team has used it yet. That is what makes coming in now worth something: what you ask for decides which nodes get built next.
What is missing, said here rather than at the end of a call. No consensus estimates and no filing diff yet. For news you get the announcements the company files with the SEC itself, not a third-party headline feed.
The repository is private. It is built on open source — an MIT-licensed fork — and that is exactly all that sentence promises.

If you run a community and want to try it with them before this is a finished product, that is the conversation I most want to have.

Talk about a partnership
Inventory · today
Catalog data nodes26
Presets ready4
Presets in flight2
SourcesEDGAR · Nasdaq · FINRA
Users outside the team0
Pricing

No credits. Nothing to ration.

Metered credits teach you to hesitate before exploring — the one thing a research canvas needs you to do freely. So there aren't any.

NowAlphaFreeEvery preset, unlimited canvases, unmetered runs. While the product is young and you're helping us find its edges.Request early access
PlannedAfter alpha~$15 / monthOne flat tier, still no credits. Bring your own model keys and it costs less. Alpha users keep launch pricing for one month.Printed now so there are no surprises later.
Method

Plain answers.

Missing one? Ask us and we'll add it here.

Which companies can I research?Ones listed on US exchanges that file with the SEC. You type the ticker — AAPL, CSCO, XOM — and the filings and figures come from there. A European company only appears if it has a US-listed ADR, and for one of those some figures refuse to compute: the cover page counts ordinary shares in the home market while the price is the ADR's, and those are not the same unit. European exchanges aren't in yet.
Is this investment advice?No. We're not an investment adviser or broker-dealer, and nothing produced here is a recommendation to buy or sell. Some nodes do label a technical condition — that a price sits above its 50-day average, say — and every one of those labels is computed in code from a figure you can click through to. None of them is a view on what you should do.
Can the AI be wrong?The written parts, yes — and every claim carries the node it came from, so you can check it. The figures are a different matter: code fetches and shapes them, so the model can't reshape them.
Where does the data come from?SEC EDGAR filings and XBRL financial data; daily prices from Nasdaq's own screener; and short interest from FINRA's own published figures. Each node names its source and links to it. One node is an embedded TradingView chart, which is a third party in the page and deliberately produces no figure anything else can read.
Do I have to build the canvas myself?No — that's the point. Presets arrive built and run; editing them is optional, and a blank canvas is there if you'd rather start from nothing.
Why doesn't anything refresh on its own?Because every step is a request to a filing, and a matching set of inputs does not mean a matching answer once a trading day has passed. Steps that have gone out of date say so and offer a refresh; naming a different company is the one instruction explicit enough to re-run the steps below it without being asked twice.
What language is the app in?English. This page exists in Spanish too, but the product itself is English only — one person builds it, and two languages inside the app is the fastest way for one of them to fall behind. It has happened here once already.
Do you train on my canvases?No.

Open your first canvas.

In alpha · no credits, ever · built on open source