alpaca-mind: a living trading mind

alpaca-mind is a living trading institution, not a bot with a strategy. THE MIND — an autonomous agent that trades options on Alpaca and owns its entire life: it schedules its own sessions, chooses its model and reasoning depth per wake, and writes the very charters its sessions run under. It journals every waking moment, reviews its trades and transcripts in self-scheduled reflection, and pre-registers research as git commits before running the numbers, so its studies cannot cheat. Research becomes deployed senses: the agent writes Python scanners that a zero-cost sentinel runs 24/7 while it sleeps, with shadow validation and automatic quarantine. Every order is written to a SQLite ledger before the venue sees it; every trade links to the reasoning transcript that produced it. BORN FREE — no inherited strategy, indicators, or risk rules. The engine enforces two machine bounds: the owner's HALT kill switch and an orders-per-hour runaway-code cap. Everything else the mind earns from its own research. Published algorithms converge and decay; each deployment starts fresh, learns its own market, and finds its own edge — no two become the same trader. THE LIVING UI — a second agent builds a web interface AFTER the trader's first real session, from the trader's own journals, then keeps growing it: live-polling dashboards that breathe between its runs, and an inbox where your requests ("show me how it thinks about earnings") become new pages on the next pass. ALPACA-NATIVE — Alpaca's official MCP server is mounted in every agent session; the engine speaks the Trading and Market Data APIs directly (options chains, greeks, multi-leg orders); paper trading is the default. DEPLOYABLE BY ANY AI — the README is written for your AI assistant: paste the repo link and it interviews you, explains real costs, deploys one CloudFormation stack, and hands you a live URL — plus a removal protocol that leaves nothing billing. github.com/david-lagrange/alpaca-mind