Hey Product Hunt 👋

I've been building trading-ops for a few months and I'm excited to finally share it. Here's the full picture.

The problem

Most trading tools give you data. Fewer give you structure. Almost none give you an audit trail that lets you see how your analysis evolved — and whether your setups actually worked.

I wanted a workflow that was:

Consistent — the same checklist applies every time, every asset class, no improvisation
Auditable — every scan dated and archived; every rescan delta-compares against the prior
Improvable — the framework is version-controlled Markdown; it gets better over time, and you control every change

What it does

trading-ops is a workspace that runs inside Claude Code — Anthropic's CLI for agentic coding workflows. You run a slash command; Claude applies a framework to live data and saves a structured Markdown analysis.

Five asset classes, one command:

/scan-macro        → regime quadrant: Goldilocks / Reflation / Stagflation / Risk-Off
/scan AAPL         → 6-pillar fundamentals + Volume Profile + VWAP + positioning + trade plan
/scan BTCUSDT      → F&G + ETF flows + perp funding/OI + liquidation heatmap + on-chain
/scan SPX          → gamma exposure + VIX term structure + AAII + breadth
/scan EURUSD       → CFTC COT + retail sentiment + rate differentials
/discover          → Finviz screener in natural language, anchored to the current regime

The symbol auto-detects its asset class — same command works for everything.

What a verdict looks like

Every scan ends with a verb (LONG / SHORT / WAIT / SKIP), an ASCII price ladder with every level sourced, and a trade table with Entry / Stop / T1–T3 / R:R / Sizing per horizon:

              BTCUSDT  -  2026-04-29  -  spot $76,335
              ==========================================

              UPSIDE  (bull resolution path)
   $84,000   ===   [T2]  heatmap cluster (forced-cover magnet)
   $80,000   ===   [T1]  heatmap density + May 30 max pain
   $79,500         -->   LONG entry  (Swing-B)
   $79,200   ===    ^^   trigger: daily close ≥ + green vector
   - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
>> $76,335   ***   CURRENT   ***
   - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
              DOWNSIDE  (bear resolution path)
   $73,800         -->   SHORT entry  (Swing-A)
   $66,000   ===   [T3]  long-dated max pain (Sept/Dec OPEX)

If there's no setup, the verdict is WAIT — and the system says so explicitly. No prose triggers, no ambiguity.

Three horizons per name

Every scan produces three independent verdicts — Positional (weeks–months), Swing (3–15 days), Day (intraday). A positional Skip is compatible with a tradeable Swing around a catalyst.

Tiers: Top Pick (deploy on trigger) / Watchlist (live spec, act on conditions) / Bench (watching, no trigger) / Skip (auto-reject or hostile)

The framework (what the LLM applies)

All intelligence lives in four read-only documents in docs/. The LLM applies them — it doesn't improvise.

macro.md — 4-regime model (growth × inflation quadrants), indicator list, FOMC/CPI/NFP playbook. Every downstream scan is gated by the current regime — a great chart in a Risk-Off regime gets Bench, not a trade.
long-term-investing.md — 6-pillar scorecard: Quality (ROIC vs WACC, margin trend), Growth (3-quarter trajectory), Valuation (P/E vs history, PEG), Balance Sheet (leverage, runway), Capital Allocation (buyback quality, M&A), Smart Money (13F changes, Form 4 insider clusters, short interest). Auto-reject rules cut obvious misses before the full analysis runs.
volume-profile.md — POC / VAH / VAL / HVN / LVN, profile shapes (D / P / b / B), naked POCs as forward targets.
vwap.md — Session / weekly / monthly / quarterly / anchored VWAP, ±1σ / ±2σ bands, reclaim/lose patterns.

Add your own framework knowledge via /ingest input/your-paper.pdf — the command shows a write plan, waits for your consent, and extends docs/ without touching existing protocol.

The self-improvement loop

After every scan, Claude audits the process itself — broken data source, wrong filter code, ambiguous step, missing edge case. It surfaces fixes with a concrete file path and proposed change. It waits for consent before editing guide/ or .claude/commands/. It never touches docs/.

Over time, the protocol gets better without losing control of what changed.

Data pipeline — no paid subscriptions required

Nine Python scripts (yfinance, SEC EDGAR, FRED, CoinGecko, Google News RSS) pre-compute data — no paid APIs required. Everything works keyless; optional free keys unlock more.

Extensible via MCPs

The base layer is research only — intentionally. But Claude Code is MCP-native, so the stack is open. Examples:

Broker MCP (Alpaca, CCXT for Weex/Binance/Bybit) — execute directly from a verdict
Chrome DevTools MCP — auto-pull TradingView charts into every scan
Playwright MCP — scrape any page without an API (broker portals, paywalled flow data)
Slack / Telegram MCP — post verdict deltas to a channel when a rescan flips
Hermes (Nous Research) — run scans 24/7 on a $5 VPS, push alerts to Telegram or Discord

Every integration is additive. The research framework stays the anchor.

Completely free

MIT license
No paid data subscriptions
Claude Code free tier works
Python 3.10+

GitHub: https://github.com/l3lackcurtains/trading-ops

Would love to hear how others use it and what protocols you'd add. The framework is designed to grow — drop a comment if you want to talk architecture.

Trading Ops - Framework-driven trading workspace in Claude Code

Replies