Twenty-three years of minute-by-minute prices for every US stock. Your agent asks a question, pays a fraction of a cent, and has its answer in milliseconds. No API keys. No subscriptions. No one filling out a signup form.
Traditional data vendors were built for humans — contracts, sales calls, monthly minimums. Cabrini is built for software that buys what it needs, when it needs it, using the x402 payment standard.
Your agent requests any ticker and date — Apple on the day the iPhone launched, or every symbol that traded last Tuesday.
The response quotes an exact price. The agent signs a USDC micropayment on Base and retries — no account, no invoice, no human.
Around 390 one-minute bars come back — open, high, low, close, volume, and trade count for every minute of the session.
Every price is quoted up front in the response itself. No tiers, no seats, no minimum spend — a research run that costs a dollar today costs nothing tomorrow if you don't run it.
This isn't a sample set. It's the full US equity market at one-minute resolution, from September 2003 through the present, cleaned and refreshed on an ongoing basis.
Large caps, small caps, ETFs, and the long tail — everything that printed a trade, including symbols that have since delisted.
Open, high, low, close, volume, and transaction count for each minute, with nanosecond-epoch timestamps.
The 2008 crisis, the flash crash, the meme-stock era — 23 years of market behavior to train on and test against.
Gap-checked and normalized to one schema across the entire archive, so 2003 parses exactly like 2026.
Cabrini speaks the protocols agents already understand — connect over MCP, discover via A2A or llms.txt, pay with x402. Everything technical lives in one place.