Skip to content

tmcga/alpha-stack

BranchesTags

Repository files navigation

Alpha Stack

CI Python 3.10+ License: MIT Skills Tools

What if every analyst had a senior MD's frameworks, a quant's toolkit, and a risk officer's discipline — loaded into their terminal and their spreadsheet?

Alpha Stack is an installable AI skill system for finance. 47 skills, 26 computational tools, a Bloomberg-inspired terminal, an Excel add-in with model templates, and a personal knowledge base that compounds over time.

Why Alpha Stack?

  • Not prompts — pipelines. Each skill is a 200-870 line execution pipeline with phased workflows, decision gates, and quality checks. Not a list of questions to ask.
  • Real math, not vibes. 26 Python tools that run DCF, Black-Scholes, Monte Carlo, Black-Litterman, LBO, DSCR sizing, equity waterfalls, and more. Every recommendation is backed by a calculation you can verify.
  • Works where you work. Bloomberg-style terminal (TUI), Excel sidebar with model templates, Claude Code skills, Claude Desktop MCP tools. Pick your interface.
  • Gets smarter over time. Personal wiki tracks every company, thesis, and methodology preference. The Idea Map surfaces thesis drift, research gaps, and actionable signals.
  • Zero dependencies. All core tools are stdlib-only Python. Install in 30 seconds. No API keys, no accounts, no subscriptions.

Install (30 seconds)

git clone https://github.com/tmcga/alpha-stack.git ~/alpha-stack
cd ~/alpha-stack && ./setup.sh

Open Claude Code. Type /sell-side. You're running.

For Claude Desktop (no terminal needed): run ./setup-mcp.sh instead. See MCP Server below.

The Workflow

Every analysis follows six phases:

  Source ──> Diligence ──> Model ──> Stress ──> Decide ──> Monitor
    │           │            │          │          │          │
  Frame the   Research    Run the    Break the   Ship the   Track the
  question    deeply      numbers    thesis      memo       position

Source — Identify the opportunity. Frame the analytical question. Diligence — Deep research using desk-specific prompt frameworks. Model — Build the quantitative case. Run Python tools, sensitivity tables. Stress — Pre-mortem. What kills this thesis? Who is on the other side? Decide — Recommendation with conviction level, sizing, and risk/reward. Monitor — Catalyst calendar, thesis-drift triggers, exit criteria.

47 Skills

Deal & Banking

Command What It Does
/sell-side Sell-side M&A — teaser, CIM, buyer mapping, bid evaluation, fairness opinion
/buy-side Buy-side acquisition — target screening, valuation, synergies, offer structuring
/lbo LBO modeling — debt sizing, returns attribution, sensitivity, exit analysis
/restructuring Distressed — liquidity triage, waterfall, fulcrum security, plan of reorg
/ipo IPO analysis — readiness, valuation, bookbuilding, pricing, stabilization
/pitch-deck Pitch deck builder — startup fundraise, deal marketing, fund pitch, internal
/investment-memo IC memo — equity L/S, PE/VC, credit, and real estate modes

Trading & Derivatives

Command What It Does
/trade Execution analysis — block trades, VWAP/TWAP, market impact, TCA
/derivatives Options pricing — Greeks, implied vol, strategy construction, convertibles
/market-making Avellaneda-Stoikov — optimal quoting, inventory management, regime adaptation

Hedge Funds

Command What It Does
/long-short L/S equity — variant perception, catalyst mapping, Kelly sizing, monitoring
/macro Global macro — regime identification, cross-asset expression, geopolitical risk
/merger-arb Event-driven — deal spread, implied probability, collar/CVR, portfolio exposure
/credit Credit analysis — Z-Score, Merton model, recovery, relative value, IG/HY/distressed

Equity Research

Command What It Does
/equity-research Initiating coverage — earnings models, trading comps, DCF, SOTP, price targets
/earnings Earnings analysis — preview, post-earnings, estimate revisions, guidance tracking

Portfolio & Risk

Command What It Does
/portfolio Portfolio construction — Black-Litterman, risk parity, factor, constrained optimization
/risk Risk analytics — VaR/CVaR, Monte Carlo stress testing, factor decomposition, tail risk
/attribution Performance attribution — Brinson-Fachler, factor, currency, fixed income, multi-period

Real Estate

Command What It Does
/re-acquisitions Property acquisition underwriting — core, value-add, opportunistic strategies
/re-development Ground-up development — cost build, lease-up, yield on cost, construction financing
/re-debt Capital stack structuring — DSCR/LTV/debt yield sizing, mezzanine, preferred, bridge
/re-reit Public REIT analysis — NAV, FFO/AFFO, implied cap rates, sector comps

Private Capital

Command What It Does
/pe-buyout Control buyouts — LBO, platform + bolt-on, returns attribution, value creation
/pe-growth Growth equity — minority stakes, unit economics, path to profitability, governance
/private-credit Direct lending — unitranche, mezzanine, covenant design, risk-adjusted yield
/secondaries LP secondaries — GP-led continuation, NAV lending, fund restructuring

Venture Capital

Command What It Does
/vc-early Pre-seed through Series A — term sheets, SAFEs, cap tables, dilution modeling
/vc-growth Series B+ — growth metrics, biotech rNPV, crypto token economics, secondary shares
/vc-fund Fund construction — portfolio math, reserves, power law, J-curve, LP reporting

Wealth Management

Command What It Does
/retirement Retirement planning — Monte Carlo, withdrawal strategy, Social Security, Roth conversion
/estate Estate & tax planning — trusts, GRATs, gifting, dynasty trusts, family office
/insurance Insurance analysis — life, disability, LTC, annuities, key person, buy-sell

Quant

Command What It Does
/quant Strategy development — signals, backtesting, overfitting detection, LLM sentiment, regime

CFO & Corporate Finance

Command What It Does
/budget Annual budget build, variance analysis, zero-based budgeting, departmental review
/forecast Rolling forecasts, 13-week cash flow, scenario planning, revenue modeling
/board-deck Board reporting, KPI dashboards, investor updates, earnings prep
/fpa FP&A — unit economics, SaaS metrics, headcount modeling, strategic finance

Accounting & Data

Command What It Does
/accounting Journal entries, accruals, closing process, chart of accounts, ASC 606/842
/financial-statements P&L, balance sheet, cash flow (indirect method), ratio analysis, DuPont
/audit Audit planning, risk assessment, substantive testing, controls, sampling
/data-entry Financial data extraction, cleaning, normalization, validation

Knowledge Base

Command What It Does
/wiki Personal knowledge base — ingest analyses, query prior research, track methodology preferences. Compounds over time.

Each skill is an execution pipeline with phased workflows, decision gates, tool integration, quality gates, hard constraints, and common pitfalls.

Every skill has a ready-to-paste example in examples/ — realistic scenarios with all inputs pre-filled so you can try any skill immediately without inventing data.

26 Computational Tools

Standalone Python calculators. Zero dependencies — just Python 3.10+. Run from CLI or import as modules.

Domain Tools
Valuation dcf.py lbo.py wacc.py irr.py
Options black_scholes.py implied_vol.py convertible.py
Fixed Income & Credit bond_yield.py merton_model.py credit_spread.py
Portfolio & Risk portfolio_risk.py kelly.py brinson.py black_litterman.py monte_carlo.py
M&A merger_arb.py
Real Estate cap_rate.py re_debt.py re_waterfall.py re_development.py re_noi.py
VC / Lending vc_returns.py loan_amort.py
Tax & Depreciation depreciation.py
Quant Trading market_maker.py
Knowledge Base wiki.py 
# Run any tool directly
python3 tools/dcf.py --fcf 100,110,121,133,146 --wacc 0.10 --terminal-growth 0.025 --shares 100
python3 tools/re_debt.py --noi 1800000 --value 28000000 --rate 0.0625 --min-dscr 1.25 --max-ltv 0.65
python3 tools/re_waterfall.py --equity 10000000 --cfs 800000,900000,1000000,1100000,15000000
python3 tools/irr.py --cfs="-1000000,200000,300000,400000,500000"

Full tool documentation at tools/README.md.

Personal Knowledge Base (/wiki)

Alpha Stack gets smarter the longer you use it. The /wiki skill builds a personal finance knowledge base — a persistent, local collection of markdown files that captures every company you research, every thesis you form, and every methodology preference you express.

Inspired by Karpathy's LLM Wiki pattern.

You:    /wiki init
Claude: Wiki initialized at ~/.alpha-stack/wiki/

You:    /wiki — ingest my AAPL analysis: services mix shifting from 22% to 35%,
        driving P/E expansion from 28x to 32x. Target $245. Key risk: EU antitrust.
Claude: Created entities/aapl.md, journal/aapl-services-thesis-2026-04-05.md
        Updated index.md with cross-references.

        --- 3 months later ---

You:    What do I know about Apple?
Claude: Per your analysis from 2026-04-05 (journal/aapl-services-thesis-2026-04-05.md):
        Thesis: services mix shift → P/E expansion to 32x. Target $245.
        Key risk: EU antitrust forcing App Store fee reductions.

How it works:

  • Entities — company/sector pages that accumulate across analyses
  • Journal — dated analysis records with thesis, findings, recommendation, and outcome tracking
  • Playbooks — your preferred methodologies ("I use 10% WACC for mid-cap tech")
  • Schema — conventions that evolve as the LLM learns your style

Everything stays local at ~/.alpha-stack/wiki/. Nothing is committed to the repo. The wiki is yours — just markdown files you can browse, git-track separately, or back up however you want.

Bloomberg Terminal (TUI)

A full terminal interface with live market data, inspired by the Bloomberg Terminal.

./setup-tui.sh && python -m tui

4 screens, keyboard-driven (F1-F4):

Screen Key What It Shows
Dashboard F1 Live market indices (S&P, Dow, Nasdaq, VIX), treasury yield curve, wiki status, recent sessions
Wiki Browser F2 Browse entities/playbooks/journal with tree nav + markdown viewer + search
Tool Runner F3 Run any of 25 tools with dynamic parameter forms, formatted results, save to session
Idea Map F4 Intelligence layer — surfaces thesis drift, outcome tracking, research gaps, risk monitors from your wiki

The Idea Map is what makes the wiki compound — it mines your knowledge base and cross-references with live market data to tell you what needs attention: "AAPL is +18% past your $245 target", "CyberVault LBO thesis is 90 days old with Open outcome", "Your DCF playbook hasn't been updated in 6 months."

Dark navy/amber theme. Live yfinance data. All data stays local.

Excel Add-in

Alpha Stack inside Excel — for bankers, PE associates, CFOs, and real estate professionals who live in spreadsheets.

./plugin/setup-excel.sh
# In Excel: Insert → Get Add-ins → Upload manifest.xml

6 sidebar tabs:

Tab What It Does
Tools Pick any of 25 tools, fill dynamic forms, run, write results to cells with banker formatting
Sensitivity Select output + two params → generates 5x5 sensitivity grid with conditional formatting
Scenarios Define Base/Bull/Bear, toggle assumptions, compare side-by-side
Templates 9 model templates with yellow input cells — LBO, DCF, Merger, Debt Sizing, Waterfall, Cap Table, Portfolio, Retirement, Operating
Audit Scan for hardcoded values, broken refs, formula errors, sign convention issues
Library Version snapshots with diff, saved sessions, wiki search

Model templates generate multi-tab workbooks with banker-grade formatting: yellow-shaded inputs, blue hardcoded values, gray labels, proper number formats ($X.XM, X.X%, X.Xx), and borders. They match the formatting conventions of every major bank and PE shop.

Claude Desktop (MCP Server)

Alpha Stack is also a finance MCP server — 45 tools that Claude Desktop can call natively. No terminal, no Python knowledge. Just ask a question and Claude runs the math.

Setup

git clone https://github.com/tmcga/alpha-stack.git ~/alpha-stack
cd ~/alpha-stack && ./setup-mcp.sh

Restart Claude Desktop. Look for the hammer icon — that means the tools are live.

Manual setup (without installer)
  1. Install uv: curl -LsSf https://astral.sh/uv/install.sh | sh
  2. Install dependencies: cd ~/alpha-stack && uv sync
  3. Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
  "mcpServers": {
    "alpha-stack": {
      "command": "uv",
      "args": ["run", "--directory", "/absolute/path/to/alpha-stack", "python", "mcp_server.py"]
    }
  }
}
  1. Restart Claude Desktop.

What It Looks Like

You:    I'm looking at a company with EBITDA of $100M. Entry at 10x, exit at 11x,
        5x leverage, 6% rate, 8% EBITDA growth, 5 year hold. What are sponsor returns?

Claude: [calls lbo_returns tool]

        MOIC: 2.99x
        IRR: 24.5%

        Returns Attribution:
          EBITDA Growth:    0.9%
          Multiple Change: 27.6%
          Deleveraging:    71.5%

You:    Size a senior loan on a $28M property with $1.8M NOI at 6.25%.

Claude: [calls re_debt_sizing tool]

        Binding Constraint: LTV
        Max Loan:     $18,200,000
        Actual DSCR:  1.34x
        Debt Yield:   9.89%
        Positive Leverage: YES

See It In Action

Full annotated example outputs with real tool results:

  • Sell-Side M&A — DCF valuation, LBO floor, buyer universe, risk matrix for a specialty chemical company
  • LBO Analysis — Returns attribution, sensitivity tables, debt schedule for a cloud security buyout
  • Long/Short Equity — Variant perception thesis, Kelly sizing, catalyst calendar for semi equipment
  • Retirement Planning — Monte Carlo simulation, ruin probability, portfolio recommendations
  • Performance Attribution — Brinson-Fachler decomposition, CIO commentary

Every skill also has a ready-to-paste example scenario in examples/ — try any skill immediately without inventing data.

Philosophy

Alpha Stack is built on seven principles documented in ETHOS.md:

  1. Assumptions are the model — every number is a bet; make the bets explicit
  2. Second-order thinking — "and then what?" is the most important question
  3. Adversarial by default — pre-mortem every thesis before it gets capital
  4. Precision over confidence — ranges over point estimates, distributions over averages
  5. Tools verify intuition — run the math, then interpret
  6. Cross-desk fluency — seeing all perspectives is the edge
  7. No hallucinated numbers — if the data doesn't exist, ask for it

Project Structure

alpha-stack/
├── CLAUDE.md              AI-native instructions for Claude Code
├── ETHOS.md               The Alpha Edge — finance AI philosophy
├── setup.sh               Claude Code skill installer (30 seconds)
├── setup-mcp.sh           Claude Desktop MCP installer
├── setup-tui.sh           Bloomberg Terminal TUI installer
├── mcp_server.py          MCP server (45 tools for Claude Desktop)
├── pyproject.toml         Python dependencies
├─��� skills/                47 skill directories (SKILL.md + prompts/)
├── tools/                 26 Python calculators (stdlib-only)
├── tui/                   Bloomberg Terminal TUI (Textual)
│   ├── app.py             Main app — F1-F4 screen routing
│   ├── bloomberg.tcss     Dark navy/amber theme
│   ├── ideas.py           Idea engine — mines wiki for actionable signals
│   ├── data.py            Live market data via yfinance
│   ├── registry.py        25 tools with full parameter metadata
│   ├── screens/           Dashboard, Wiki Browser, Tool Runner, Idea Map
│   └── widgets/           Command bar, ticker bar, forms, result panel
├── plugin/                Excel Office Add-in
│   ├── server/            FastAPI server wrapping all tools as REST endpoints
│   │   ├── app.py         API server with CORS + static file serving
│   │   ├── templates.py   9 model template definitions
│   │   └── output_mapper.py  Tool results → Excel cell layouts
│   ├── addin/             Sidebar UI (HTML/CSS/JS + Office.js)
│   │   ├── manifest.xml   Office Add-in manifest for sideloading
│   │   ├── taskpane.html  6-tab sidebar (Tools, Sens, Scenarios, Templates, Audit, Library)
│   │   ├── taskpane.css   Professional banker-grade theme
│   │   └── taskpane.js    Client logic — forms, API calls, cell writes
│   └── setup-excel.sh     One-script installer
├── examples/              26 ready-to-paste scenarios + 5 annotated outputs
├── tests/                 113 pytest tests
└── docs/                  Workflow documentation

Contributing

See CONTRIBUTING.md for guidelines on adding new skills, prompts, or tools. Scaffolding scripts in scripts/ create properly structured files from templates. All contributions welcome.

License

MIT — use it however you want.

About

Alpha Stack is an installable AI skill system for finance. 47 skills, 26 computational tools, a Bloomberg-inspired terminal, an Excel add-in with model templates, and a personal knowledge base that compounds over time.

Resources

Readme

License

MIT license

Contributing

Contributing
Activity

Stars

2 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages