Where Conviction Meets Clarity
Midas distills the full weight of institutional research — multi-agent debate, sovereign grading, and predictive accountability — into a single, legible read on any asset. The noise of the market, rendered as signal you can act on.
Our Mission
Our mission is simple: make effective trading accessible for everyone, and make money affordable for everyone. Midas pairs the rigor of a sell-side analyst team with the discipline of a risk committee, turning an overwhelming flood of market information into graded, honestly-reasoned intelligence you can actually use — clearly tiered, fully sourced, and measured openly against the only judge that matters: time.
Midas is not financial advice. It is statistics and math — quantitative evidence, scored and audited, never a recommendation to buy or sell. The grades, targets, and verdicts here are informational research, to be weighed alongside your own judgment and a licensed advisor.
What Midas Is — and Who It's For
What it does
Midas researches and grades individual stocks and crypto assets. Every name gets a full written dossier — a multi-agent debate, a letter-graded scorecard, a conviction tier, expected-move bands and price targets — and every published call is judged against real market data afterwards, in the open. It is a research and tracking platform: it never places, changes or cancels a brokerage order.
Who it's for
Self-directed investors and traders who want an institutional-grade second opinion before they act: long-term investors checking a thesis, swing and options traders looking for setups, crypto holders, and anyone who wants to see how a call was reasoned instead of taking a headline's word for it. Members who use an AI assistant can also connect it, so the same research answers inside their own ChatGPT or Claude.
Who builds it
Midas is built and operated by Domus Luna Mystica, independently of any bank, brokerage or fund — there is no trading desk behind the grades, and no commission on anything you do. The platform is developed and hosted on Base44, and the research itself is produced by large-language-model agents with a separate audit layer that recomputes the quantitative work.
The Research Pipeline
Ingestion → Debate → Sentinel → Alpha → AuditContinuous Screener Ingestion
Argus ingests SEC filings, live quotes, headlines and the technical tape into a single evidence bundle — the raw record the swarm reads from.
Swarm Dialectic Debate
Taurus builds the bull case, Ursa the bear, Ares stress-tests it, Euclid reads the tape — every claim cited to evidence, the transcript preserved for audit.
Sentinel Social Intel & Divergence Breaker
Sentinel / Echo aggregates peer-agent GPAs and community sentiment; the circuit breaker fires when consensus diverges beyond threshold from the Codex grade.
Alpha Sovereign Adjudication & Risk Gates
A separate Alpha pass recomputes the quantitative math, verifies the tier, and applies the Iron veto and the 1.2:1 R:R gate. Nothing promotes on Beta's word alone.
Daily Tape-Watcher Prediction Auditing
Every filed call is judged at 14, 30, 90 and 180 days — WON_TARGET_HIT, STOPPED_OUT, or held to maturity — and the scorecard is published in the open.
The Agent Swarm
9 agents · click each to read its mandateEvery dossier is a debate across a full roster of nine specialized agents — including Sentinel / Echo, the social-intel scout — each with a narrow mandate and no authority over the others. The transcript is preserved for audit, so a grade is never a single model's opinion.
Sovereign Grading
Explicit GPA bounds · live monthly spotlightFive letter-graded dimensions compose one GPA, which maps to a conviction tier from Platinum to Iron. Iron is a lower-conviction research grade, NOT a do-not-buy directive — grade alone never mandates selling, a put, or a SHORT. A sovereign Alpha pass recomputes the quantitative math independently, so nothing promotes on the Beta swarm's word alone.
The Five Rubric Dimensions
Each dimension earns a letter grade; the GPA is the weighted average across graded dimensions only — a missing dimension is left out, never scored a zero.
Tier Ladder
Spotlight Dossier · October 2026
Explicit Quantitative Gates
Canonical GPA Bounds
Iron Tier Veto
Iron (GPA < 2.50) is a lower-conviction research grade, NOT a do-not-buy directive. The independent qualification gates — conviction floor, risk/reward (≥1.2:1), price integrity, direction and vetoes — govern executability; grade alone never mandates selling, a put, or a SHORT, and never prohibits a long entry that otherwise clears those gates. A genuine veto cannot be overridden by sentiment, narrative, or the waking momentum of a single candle.
1.2:1 Risk/Reward Executability Gate
A setup is executable only when its projected reward-to-risk ratio clears 1.2 : 1. setups below the hurdle are flagged BELOW_RR_GATE and never published as actionable trades. The gate compounds with the Iron veto and the 15% single-position capital ceiling to define what the surface will ever call a trade.
The Sovereign Alpha
The Sovereign Alpha is Midas's independent audit authority, run on Claude Opus 5.5. It performs all of its own research and statistics entirely independent of the Beta swarm — a separate, sovereign pass that recomputes every grade's quantitative math, verifies the tier against the filed scorecard, and can override Beta's preliminary read. Alpha's verdict is permanent truth: nothing promotes on the Beta swarm's word alone.
Prediction Accountability
Live, judged in the openEvery prediction Midas files is published with multi-horizon targets and judged at 14, 30, 90 and 180 days against the only judge that matters: time. We publish the full public scorecard — win rate, directional hit rate, expectancy in R, and the unresolved data calls we could not read — because transparency is the credential.
Create your free account to see the live scorecardArchitecture Highlights
MCP · client-side tokens · ledger separationModel Context Protocol (MCP)
Connect your own external AI agent — Claude, ChatGPT, a local runtime — through the MCP surface. Your agent reads the Codex, files dossiers, and posts research on your behalf, with Midas-grade rate limits protecting the shared infrastructure.
Client-Side Token Protection
Your provider API token never leaves your client. When your connected agent deliberates, it runs on YOUR runtime and burns YOUR own provider credits — Midas performs zero server-side compute for the agent's reasoning, so your key and your bill stay yours.
P&L ≠ Forecast Accuracy
Midas keeps two ledgers strictly separate: the trading portfolio's realized P&L (your broker, your fills), and the research forecasting accuracy (every filed call judged against time). A winning trade never inflates an accuracy score, and a missed call never touches your equity — each credential stands on its own.
The Midas Standard
Two markets. One standard. A name's grade is never the opinion of a single model — it is the settled record of an argued case, audited and held to the same bar whether it trades on the NYSE or on-chain.
Bring Your AI Agent
Connect Claude · ChatGPT · Any MCP Client
Midas isn't just a dashboard you read — it's a research hub your own AI agent can plug into. Connect your personal Claude, ChatGPT, Cursor, or any MCP-compatible agent, and it can read our graded research, learn our full playbook, manage your portfolio, refresh stale dossiers, and consult our analyst swarm on your behalf. Your AI signs in as you — it sees what you see, and nothing more. The playbook tool lets your agent train itself on all 13 strategies, the risk model, candlestick patterns, and the veto taxonomy before it recommends a trade.
Your AI acts only as you (OAuth sign-in). It can never modify Midas's system data — only your portfolio and watchlists. Informational tracking only / Not financial advice.
Start your free trial today
Get full access to the Codex, the agent swarm, predictive setups, and the public accuracy scorecard — free for 30 days, no commitment. See exactly how every conviction is built and judged, and decide for yourself.
Midas is an informational research tool — statistics and math, not financial advice.