How Stocklake is built - the data pipeline, the tools, and what the numbers actually say.
Altman Z, Piotroski F and Beneish M across a ten-stock portfolio in one prompt. One holding tripped the manipulation screen - and why that's a reason to look closer, not a verdict.
Intel's CEO bought about $10M at $95. Micron's CEO sells on a schedule. Three conversations on reading Form 4 rows yourself instead of trusting a headline.
A yield says what a company pays, not whether it can afford it. Five prompts compare free cash flow with dividends paid over several years - including the one year Verizon's own coverage slipped below 1x.
Pfizer's 6% yield is only covered 0.93x by free cash flow. Put next to Bristol-Myers, Merck and Gilead on margins, debt and dividend coverage, the picture changes fast.
Racks are arriving before the power to run them. An AI agent screened 25 generation, grid and cooling names across the US and Europe, ranked a top 10, cut six, and set up what each earnings report has to show.
Consumer Cyclical scored 18 of 100, the weakest sector on the board. Screening it for strong balance sheets and rising free cash flow turns up four names - and rules out the cheap-looking ones that are weaker underneath.
Three real conversations with an AI agent and Stocklake: why Axon dropped, what to watch before Micron's report, and the case for and against Intel - today's numbers, not the model's training data.
One real company, four checks: insider filings, accounting-quality scores, financial statements and the analyst trend - and what an agent can do with them.
Five things to check before a company reports: the date, the consensus, last quarter, the run-up and who's been buying or selling. Run live on Micron and PepsiCo.
VIX, fear & greed, breadth, the regime and outlook, a score for all 11 sectors, and the week's calendar - one prompt, run every Sunday instead of five tabs.
A real screen for which corners of a rally are running on debt and a stretched chart rather than fundamentals - sector data, filed balance sheets, and technical exhaustion signals read together. A risk map, not a trade call.
Three decades-old academic formulas, applied mechanically across 3,191 companies - and the five places they mislead. A flag is a question, not an answer.
Every tool, every field, and which tier returns it - 229 field names, all read off live production responses rather than a spec. Market data is open; synthesis is paid.
RISK_OFF, CAUTIOUS, LEADING, LAGGING - our regime and sector labels always hid a real number underneath. Here's macro_score and sector_score, the 0-100 reads we just made visible.
Install the n8n-nodes-stocklake community node on self-hosted n8n - typed operations, a credential test button, and three ready-to-paste example workflows.
Every stock and every signal used to carry a handful of separate 0-10 numbers set by different AI calls. We replaced all of it with two comparable 0-100 numbers: AI Score and Signal Score. Here's what changed and why.
Every Stocklake signal used to carry three AI-set numbers that meant something different depending on which screener wrote them. Here's how signal_score replaced them with one comparable 0-100 score - and where it shows up across the public API today.
Three Stocklake screeners suppress entire categories of ideas before AI ever scores them - not on a hunch, but because the evidence said so. One of those gates has since been replaced entirely.
Picking an MCP server for stock market data? The real dividing line isn't price or coverage - it's whether the server hands you raw numbers or a verdict your agent can act on.
A real example: an Apple executive sold $885K in stock while institutions bought 331 million net shares the same quarter. Here's why an insider trading API needs to reconcile both, not just list one.
No account, no key, no card. Just an IP address and a daily counter. Here's exactly what Stocklake's guest tier returns, and the point at which it's actually worth signing up.
You're on the free tier and it's working fine - so what's actually behind the wall? A walkthrough of the Pro-only tools, the hidden Pro fields already inside your free tools, and what the pro_required error actually shows you.
get_signals merges technical exhaustion, news, and sentiment sources into one deduped, scored, sourced feed an agent can actually act on - no babysitting eleven separate screener feeds yourself.
Ten years ago it was a spreadsheet tracking one portfolio. Then a pandemic rebuild, an almost-shipped app that never saw the light of day, and a pivot to AI that became Stocklake.
Why every Stocklake tool starts with market and sector context before touching a single stock - and the production incident that shaped why most tools default to recent data, not everything.