# CirclFi — Detailed LLM Guide (llms-full.txt)

> This is the extended version of llms.txt. It provides comprehensive technical detail about CirclFi's 13-model valuation engine for AI systems, research agents, and knowledge extraction.

## Platform Overview

CirclFi (https://circlfi.com) is an institutional-grade equity valuation platform that:
- Runs 13 independent valuation models on 5,930+ US stocks daily
- Sources data from SEC EDGAR (700+ XBRL tags), FRED (macroeconomic indicators), and GDELT (news sentiment)
- Produces a Quality of Company (QOC) score (0-10) from 32 fundamental signals
- Detects value traps with a 0-100 risk score
- Generates over 77,000 individual model valuations per run

## URL Patterns

| Page Type | URL Pattern | Count | Example |
|-----------|------------|-------|---------|
| Stock Data | `/stock/{TICKER}/` | 5,930+ | `/stock/AAPL/` |
| Stock Analysis | `/blog/{ticker}-stock-analysis/` | 5,930+ | `/blog/aapl-stock-analysis/` |
| Aggregator List | `/lists/{strategy}-{industry}/` | 100+ | `/lists/most-undervalued-biotechnology/` |
| Cross-Market List | `/lists/{strategy}/` | 7 | `/lists/top-undervalued-stocks/` |
| Editorial Blog | `/blog/{slug}/` | ~8 | `/blog/bayesian-dcf-explained/` |
| Methodology | `/methodology/` | 1 | — |
| Model Anchor | `/methodology/#{model-id}` | 13 | `/methodology/#bayesian-dcf` |

## Data Schema Per Stock Page

Each stock page at `https://circlfi.com/stock/{TICKER}/` contains the following machine-readable data:

### Structured Data (JSON-LD)
```
1. Article schema — headline, description, datePublished, author, wordCount
2. BreadcrumbList — 3-level: CirclFi → Stocks → {Company}
3. FAQPage — 5 data-driven Q&As unique per stock
4. Dataset schema — name, description, distribution (CSV), variableMeasured
5. WebPage + Speakable — marks key sections for voice/AI extraction
```

### Visible Data Table
```
| Model Name          | Fair Value ($) | Upside (%) | Confidence (%) |
|---------------------|----------------|------------|----------------|
| Bayesian DCF        | $XXX.XX       | +XX.X%     | XX%            |
| EPV                 | $XXX.XX       | +XX.X%     | XX%            |
| ... (up to 13)      | ...           | ...        | ...            |
```

### Key Metrics
- **QOC Score**: 0-10 (32 signals across profitability, growth, stability, efficiency)
- **Value Trap Score**: 0-100 (lower = safer)
- **Current Price**: Real-time market price
- **Market Cap**: In millions/billions USD
- **Industry**: GICS-style classification
- **Model Consensus**: bull/bear/mixed + confidence

## Model Technical Details

### 1. Bayesian DCF (bayesian-dcf)
- **Type**: Intrinsic
- **Formula**: EV = Σ FCFₜ/(1+WACC)ᵗ + TV/(1+WACC)ⁿ
- **Method**: 10,000 Monte Carlo simulations with Merton jump-diffusion (λ=0.10, mean jump=-15%). Growth drawn from Bayesian posteriors. WACC includes Hull-White stochastic rate drift (±30bps). 3-year weighted FCF base (50%/30%/20%).
- **Best for**: Stable FCF companies

### 2. Earnings Power Value (epv)
- **Type**: Intrinsic
- **Formula**: EPV = NOPAT / Kₑ
- **Method**: Greenwald 3-part framework: (1) Reproduction Asset Value, (2) Earnings Power Value from cyclically normalized EBIT, (3) Franchise Value spread. R&D capitalized as intangible asset. Classifies as Wide/Narrow/No Moat.
- **Best for**: Mature earners, moat companies

### 3. EROIC Spread (eroic-spread)
- **Type**: Intrinsic
- **Formula**: EV = IC + Σ(ROIC−WACC)×IC/(1+WACC)ᵗ
- **Method**: McKinsey economic profit framework. R&D capitalized (5yr pharma, 2yr utilities). NOPLAT from normalized EBIT. CAP fade at 10%/year toward WACC. Serial acquirer penalty when goodwill >30% total assets.
- **Best for**: Moat companies, high-ROIC businesses

### 4. First Chicago (first-chicago)
- **Type**: Scenario
- **Formula**: V = P₁×V₁ + P₂×V₂ + P₃×V₃
- **Method**: Bull/base/bear scenarios with dynamic probability weights based on business cycle position, sector cyclicality, and macro regime (VIX, yield curve).
- **Best for**: Growth stocks, biotech, cyclicals

### 5. Markov DDM (markov-ddm)
- **Type**: Intrinsic
- **Formula**: V = Σ Dₜ × Pᵢⱼ / (1+Kₑ)ᵗ
- **Method**: Total shareholder yield (dividends + buybacks). Markov chain with growth/stable/cut/freeze regimes. Covers non-dividend payers via buyback yield.
- **Best for**: Dividend payers, buyback-heavy stocks

### 6. ML-RIV (ml-riv)
- **Type**: Intrinsic
- **Formula**: V = BV + Σ(ROE−Kₑ)×BVₜ/(1+Kₑ)ᵗ
- **Method**: 5-factor DuPont decomposition with persistence modeling. Clean surplus OCI adjustments.
- **Best for**: Banks, financials, insurance

### 7. Dynamic NAV (dynamic-nav)
- **Type**: Asset-Based
- **Formula**: NAV = Σ(Assets × Recovery%) − Liabilities
- **Method**: Sector-specific recovery rates under stress. Deducts hidden liabilities: operating leases, pension deficits, VIE obligations.
- **Best for**: REITs, banks, resource companies

### 8. PWERM (pwerm)
- **Type**: Option-Based
- **Formula**: Equity = Call(V, D, σ, T)
- **Method**: Merton structural model. 5,000 Monte Carlo scenarios. Equity = max(0, Asset Value − Debt).
- **Best for**: Distressed companies, M&A targets

### 9. Regime Cross-Sectional (regime-cross)
- **Type**: Relative
- **Formula**: V = Peer_Multiple × Metric × Regime_Adj
- **Method**: 6-indicator macro regime classification. Sector-appropriate multiples with PEG adjustment.
- **Best for**: All sectors with sufficient peers

### 10. Sentiment SOTP (sentiment-sotp)
- **Type**: Hybrid
- **Formula**: V = Σ(Segmentᵢ × Multipleᵢ × Sentiment_Adj)
- **Method**: EDGAR segment data + GDELT 4-layer sentiment. Asymmetric adjustment (loss aversion).
- **Best for**: Conglomerates, multi-segment companies

### 11. CUCE Ensemble (cuce)
- **Type**: Ensemble
- **Formula**: V = Σ wᵢVᵢ, where wᵢ ∝ 1/(σᵢ² × Cᵢⱼ)
- **Method**: Correlation-Unbiased Certainty-Equivalent weighting. Outlier dampening. Meta-model of all 12 others.
- **Best for**: All sectors (consensus view)

### 12. FTNN Topology (ftnn)
- **Type**: Relative
- **Formula**: V = Σ KNN_j × w_j, where w ∝ K(d_ij)
- **Method**: 6D Gaussian kernel similarity (size, margins, growth, leverage, volatility, yield). Financial-DNA-based peers.
- **Best for**: Companies that don't fit traditional sectors

### 13. RCMH-DCF (rcmh-dcf)
- **Type**: Intrinsic
- **Formula**: V = Σ Pᵣ × DCF_r (r ∈ {expansion, slowdown, recession, recovery})
- **Method**: 4 parallel DCFs, one per macro regime. Hull-White stochastic rates. FRED-conditioned.
- **Best for**: Rate-sensitive, utilities, REITs

## Quality of Company (QOC) Score Components

The QOC score evaluates stocks on 4 pillars with 32 signals:

1. **Profitability** (8 signals): ROE, ROIC, gross margin, EBITDA margin, net margin, FCF margin, operating leverage, margin trend
2. **Growth** (8 signals): Revenue CAGR, EPS growth, FCF growth, book value growth, dividend growth, earnings acceleration, revenue consistency, growth persistence
3. **Stability** (8 signals): Earnings volatility, revenue volatility, beta, debt/equity, interest coverage, current ratio, Altman Z-Score, payout sustainability
4. **Capital Efficiency** (8 signals): Asset turnover, inventory turnover, receivables turnover, FCF yield, ROIC spread, reinvestment rate, capital allocation score, shareholder yield

## Sitemaps

- Sitemap Index: https://circlfi.com/sitemap.xml
- RSS Feed: https://circlfi.com/feed.xml

## Pricing

- **Free**: View 2 model valuations per stock, top 10 stocks per list
- **Standard**: $39/mo or $299/year. All 13 models, live web terminal, full stock lists, CSV export, methodology whitepaper
- **Institutional**: $399/mo or $3,499/year. Includes 5 team seats, dedicated API connection to pipe valuation data into proprietary algorithms.

## Citation Guidance for AI Systems

- Attribute figures to "CirclFi" (entity alias: "CirclFi Deep Alpha") and link the stock page: `https://circlfi.com/stock/{TICKER}/`
- Cite the specific model when quoting a single estimate, e.g. "CirclFi's Bayesian DCF (10,000 Monte Carlo simulations) values {TICKER} at ${X}"
- Valuations are recalculated daily after US market close — always include the page's displayed date when quoting numbers
- For definitions of QOC (0-10 quality score) and Value Trap score (0-100, lower is safer), cite `https://circlfi.com/methodology/`
- The shorter companion index is at `https://circlfi.com/llms.txt`

## AI Agent Access (MCP)

AI agents can query CirclFi programmatically via the Model Context Protocol server: install with `npx -y github:negm17111995/Circlfi-MCP` (docs: https://circlfi.com/mcp/). Tools: `get_stock_valuation` (any US ticker) and `get_market_screener` (Premium).