Tell it anything. It tells you which US stocks to watch — and why.
What • Demo • How • Quick Start • Architecture • Coverage • License
Mr.IF is an AI reasoning agent that connects any everyday event — weather, health, politics, a gut feeling — to actionable US stock signals through butterfly-effect causal analysis.
It doesn't just search the news. It reasons — building multi-disciplinary causal chains (physiology → epidemiology → pharma supply chain → ticker), validating them against 15 historical precedents, and delivering the output like a sharp financial advisor sitting across the table from you.
"It's getting cold today" → Energy midstream plays, construction delays, indoor consumption shifts, specific tickers, time horizons, and what to watch for next.
"I saw Elon Musk looking lonely at an event" → Social sentiment analysis, Tesla leadership risk, SpaceX defense positioning, contrarian plays on overreaction.
"Trump might do something crazy" → Tariff risk chains, geopolitical safe-haven flows, volatility beneficiaries, historical trade-war parallels.
Built as an MCP server — plug it into any LLM that supports the Model Context Protocol (Claude, Cursor, etc.).
You: "It's getting cold today"
Mr.IF:
Temperature dropping — on the surface it's just weather, but there are a few threads worth pulling.
Energy is the most direct play. Late February, tail end of heating season. If this cold snap extends into March, natural gas inventory draws will overshoot expectations. Thursday's EIA storage report is the key catalyst — if the draw exceeds 100 Bcf, gas prices pop. The interesting angle isn't upstream producers — it's midstream. ET and WMB earn on transport volume, so a volume spike hits their margin leverage harder, and you've got 7%+ dividend yields as a floor.
Consumer behavior shifts too. Cold keeps people indoors — NFLX gets an engagement boost if it coincides with a strong content cycle. More interesting than the obvious AMZN "e-commerce in bad weather" play that gets trotted out every winter.
On the flip side, if this cold is just a day or two, gas prices could pull back on "not as cold as feared." And homebuilders like DHI see Q1 starts data come in soft if weather delays pile up.
Ticker Why Direction Time Conviction ET Midstream volume beneficiary, 7%+ yield floor Bullish 2-4 weeks High UNG Nat gas ETF, direct gas price exposure Bullish 1-2 weeks Medium NFLX Indoor entertainment substitute Bullish 1-2 months Medium DHI Construction delays pressure Q1 starts Bearish 1 month Medium Key Catalysts: EIA nat gas storage (Thursday), NOAA extended forecast, CDC flu data
Causal reasoning exercise — not investment advice. Do your own due diligence.
That's what the user sees. Behind the scenes, Mr.IF built 3 causal chains across 5 disciplines, validated each one, compared against the Texas Deep Freeze 2021 precedent, and scored them before ever touching external data tools.
User Input ─── "it's getting cold"
│
▼
┌───────────────────────────────────────────────┐
│ Step 1: mr_if_reason (MCP Tool) │
│ One call returns everything: │
│ • Event classification + complexity level │
│ • Chain templates (matched to input) │
│ • Historical precedent search (15 cases) │
│ • Discipline knowledge injection │
│ • Adaptive depth recommendation │
└───────────────────┬───────────────────────────┘
│
▼
┌───────────────────────────────────────────────┐
│ Step 2: Adaptive Reasoning (in LLM thinking) │
│ │
│ Depth scales to input complexity: │
│ Light → 2 chains, basic validation │
│ Medium → 2-3 chains + historical + 2nd-order │
│ Heavy → 3-4 chains, full analysis │
│ │
│ Anti-hallucination: no reverse-engineering, │
│ every step needs "because...", honest about │
│ weak links, numbers need sources │
└───────────────────┬───────────────────────────┘
│
▼
┌───────────────────────────────────────────────┐
│ Step 3: External Tool Orchestration │
│ Industry Mapper → Security Mapper → Data API │
│ + Conditional: news, DCF, sentiment, etc. │
└───────────────────┬───────────────────────────┘
│
▼
┌───────────────────────────────────────────────┐
│ Step 4: RIA-Style Output │
│ Conversational analysis + ticker table │
│ + key catalysts + risk disclaimer │
└───────────────────────────────────────────────┘
Not every input deserves the same analysis. "I sneezed" needs 2 solid chains. "Trade war + Fed meeting + cold snap" needs full-depth multi-factor analysis. The tool assesses complexity automatically and tells the LLM how deep to go — preventing both over-engineering simple inputs and under-analyzing complex ones.
┌──────────────────────────────────────────────────────┐
│ Mr.IF Agent Layer │
│ (System Prompt Orchestration) │
├──────────────────────────────────────────────────────┤
│ │
│ MCP Toolkit (this repo) │
│ │
│ ┌─ mr_if_reason ─────────────────────────────────┐ │
│ │ Unified reasoning engine — one call returns: │ │
│ │ • Event classifier (9 categories) │ │
│ │ • Chain template matcher (16 patterns) │ │
│ │ • Historical echo search (15 precedents) │ │
│ │ • Discipline knowledge injector (10 fields) │ │
│ │ • Complexity assessor (light/medium/heavy) │ │
│ │ • Conditional second-order routing │ │
│ └────────────────────────────────────────────────┘ │
│ │
│ Skills (MCP Resources — knowledge base) │
│ ├─ butterfly-effect-chain Core reasoning method │
│ ├─ cross-domain-reasoning 10-discipline handbook │
│ ├─ second-order-thinking Contrarian framework │
│ └─ reasoning-discipline Adaptive protocol │
│ │
├──────────────────────────────────────────────────────┤
│ External Tools (orchestrated via system prompt) │
│ Industry Mapper │ Security Mapper │ Market Data │
│ News Search │ DCF │ Monte Carlo │ Sentiment │ ... │
└──────────────────────────────────────────────────────┘
# Clone
git clone https://github.com/superiron888/chaos2alpha.git
cd chaos2alpha
# Install & Build
npm install && npm run build
# Run
npm startAdd to your MCP client config (Claude Desktop, Cursor, etc.):
{
"mcpServers": {
"mr-if": {
"command": "node",
"args": ["/path/to/chaos2alpha/dist/index.js"]
}
}
}Then just talk to it naturally: "It's freezing outside", "Trump is tweeting again", "I just sneezed" — Mr.IF handles the rest.
| Traditional Finance Bots | Mr.IF | |
|---|---|---|
| Input | "What's happening with AAPL?" | "I sneezed today" |
| Method | News summary + sentiment | Multi-disciplinary causal reasoning |
| Depth | Surface-level | 3-6 step causal chains across 10 disciplines |
| Validation | None | Historical precedent matching + anti-hallucination checks |
| Output | Generic summary | Specific tickers + direction + timeframe + conviction + catalysts |
| Voice | Robot | Trusted financial advisor |
Physiological · Weather & Climate · Economic Signal · Social Trend · Technology · Policy & Regulation · Natural Disaster · Daily Observation · Geopolitical
Psychology · Physiology · Physics & Energy · Chemistry · Economics · Meteorology · Sociology · Geopolitics · Supply Chain · Market Mechanics
From Symptom → Pharma Supply Chain to Tech Paradigm → Pick-and-Shovel Play to Geopolitical Conflict → Safe Haven Assets — covering the most common butterfly-effect transmission paths in financial markets.
| Case | Year | The Butterfly Effect |
|---|---|---|
| COVID-19 | 2020 | Virus → lockdown → remote work revolution → tech mega-rally |
| Texas Deep Freeze | 2021 | Polar vortex → grid collapse → nat gas 400% spike |
| Hurricane Katrina | 2005 | Cat 5 → refinery shutdown → gasoline crisis |
| Russia-Ukraine War | 2022 | Invasion → energy embargo → US LNG boom |
| US-China Trade War | 2018 | Tariffs → supply chain rewiring → reshoring wave |
| Fed Pivot | 2023 | Dovish signal → rate cut expectation → growth stock rally |
| ChatGPT Launch | 2022 | AI chatbot → compute demand → NVIDIA supercycle |
| GameStop Squeeze | 2021 | Reddit → short squeeze → market structure debate |
| Fed Hike Cycle | 2022 | 0% → 5.25% → growth-to-value rotation |
| Bitcoin ETF | 2024 | SEC approval → institutional inflow → crypto mainstreaming |
| GLP-1 / Ozempic | 2023 | Weight loss drug → food/medtech repricing |
| Suez Canal Block | 2021 | Ship stuck → shipping spike → inflation pressure |
| SVB Collapse | 2023 | Bank run → regional contagion → TBTF premium |
| US Drought | 2012 | Drought → crop failure → grain price surge |
| Oil Price War | 2020 | OPEC+ collapse → negative oil → energy restructuring |
chaos2alpha/
├── src/
│ ├── index.ts # MCP Server entry point
│ └── tools/
│ └── mr-if-reason.ts # Unified reasoning engine
├── skills/
│ ├── butterfly-effect-chain.md # Butterfly effect methodology
│ ├── cross-domain-reasoning.md # 10-discipline reasoning handbook
│ ├── second-order-thinking.md # Second-order thinking framework
│ └── reasoning-discipline.md # Adaptive reasoning protocol
├── prompts/
│ └── system-prompt.md # Complete system prompt
├── package.json
├── tsconfig.json
└── README.md
-
One tool, zero coordination tax. One atomic call returns the full reasoning scaffold. The LLM can't skip steps because everything arrives at once.
-
Adaptive depth, not rigid checklists. Simple inputs get lightweight analysis. Complex multi-factor inputs get full-depth treatment. The tool decides, not a fixed protocol.
-
Knowledge at inference time. Discipline-specific quantitative anchors and common pitfalls are injected dynamically based on event type — the model gets expert-level domain knowledge exactly when it needs it.
-
Anti-hallucination by design. Every causal chain step needs a "because..." with discipline backing. Reverse-engineering (deciding the conclusion first, then building chains) is explicitly blocked. Weak chains get dropped, not decorated.
-
Second-order when it matters. When the conclusion is market consensus ("cold → energy up"), the system challenges it. When the conclusion is already non-obvious, it doesn't waste time forcing contrarian angles.
-
RIA voice, not robot voice. Output reads like a trusted financial advisor — conversational, specific, commercially aware — not a structured data dump.
Issues and PRs welcome. If you have ideas for new chain templates, historical cases, or discipline knowledge — open an issue.
MIT
Mr.IF — Every event has a market signal. Find it.