CharmCast runs a server-side pipeline that auto-pulls the full chain across every active expiration, computes first, second, and third-order Greeks, runs a 25-layer volatility framework, and synthesizes structured trade ideas with AI-generated execution plans - entry, stop, multi-tier targets, breakeven, trail, news kill. Your job: review, ask follow-ups, decide.
The full guided tour walks you through every view - engine, analysis, orderflow, vol, journal - so you can see the layout, the AI plan card, the dealer-gamma views, and the briefings before you ever hand over an email.
From a glance at the at-a-glance dashboard, to the guided tour, to the daily macro briefing, to the under-the-hood pipeline — this is the surface area a Pro/Max account unlocks.
Skip the marketing fluff. Here's everything that runs every minute behind your tape.
Five-parameter arbitrage-free SVI per expiry. LM solver, butterfly + calendar arb checks in closed form. Denoises wing IVs feeding the RND and 25Δ-RR.
Per-strike charm × time-to-close projected forward in 30-min buckets. When projected hedge magnitude exceeds 1× ADV, FORCED-FLOW regime fires.
Per-contract Lee-Ready classifier across 15 minutes. Net signed call/put × delta = live dealer-net-Δ. Fresh-flow read the static OI-based GEX can't see.
Three online-fitted Hawkes processes (trades · big prints · sweep clusters). Branching ratio n > 0.75 fires CRITICAL - a vol-burst leading indicator.
Cont-Kukanov-Stoikov OFI extended through 5 book levels with exponential depth weights. Spoof-classified and discounted live.
Bańbura-Modugno-style nowcast from 6 high-frequency proxies - copper/gold · 5y5y · HY OAS · WTI · baltic · retail. Bucketed EXPANSION → CONTRACTION.
Geopolitical Risk index shock × empirically-calibrated Brunel-GMC betas: Energy +1.40, Defense +1.20, EM +1.10, Tech −0.60. Per-sector forward-vol bumps.
Force-directed graph linking every indicator → reason → level → setup → alarm. Click any node to see what drives it. Cross-tab cross-reference for free.
That's eight of the twenty-one. Each has a citation, each is wired into the AI prompt, each has a live UI cell. See all twenty-one →
All wired together. One screen. No spreadsheet stitching.
Delta, gamma, vega, theta, rho · vanna, charm, vomma, veta · speed, zomma, color, ultima. Per strike, every expiration.
Realized vs implied · IV rank · skew · vol cone · forward vol · tail-risk skew · implied moments · vol forecast · stochastic-vol fit · variance-swap premium · trendiness · expected move.
Confirmed setup → JSON plan with entry, multi-tier targets, stop, breakeven, trailing rules, time stop, news kill. Conviction A+ to C.
Live CVD, tape signals, dealer regime, footprint heatmap. Rejection blocks, FVG, CISD, BPR, OTE, BSL/SSL - all auto-detected.
Every trade auto-graded A+ to F by R-multiple. CSV import for Apex, TopStep, MyFunded, FTMO, Tradovate, NinjaTrader.
Voice-matched morning read. Cross-asset macro, GEX walls, long-term positioning, recent news - synthesized in 30 seconds.
Direct retail (Tradovate, IBKR, Alpaca, MT5) and every major prop firm (Apex, TopStep, FTMO, MyFunded, FundedNext, FXIFY).
Server-side risk checks before every order: max notional, daily loss, single-trade %, concurrent positions, news-window kill, consecutive-loss pause.
Trendline, ray, fib, rectangle, measure, text - pinned to time/price so they survive pan and zoom.
Server-side pipeline auto-hydrates the full chain and runs the analysis in the background. No GRAB buttons, no waiting on your machine. Walls, regime, expected move, key levels - already live when you open the tab.
The AI's structured 13-section call is ready when you arrive. Open Chat to ask "what invalidates the long?" or "what if gamma flip breaks?" - the model answers with the same context it just used.
If the read makes sense, the Executions tab turns it into a JSON plan: entry, stop, multi-tier targets, management plan, conviction grade - all enforceable by your broker.
Risk gates verify the trade server-side. Order routes to your connected broker. Journal entry is auto-created with the AI's reasoning.
The structure CharmCast reads in the background, every minute, on every symbol you're watching.
Call walls (where rallies stall), put walls (where selloffs catch), and the gamma flip line. Above flip = dealers dampen moves; below = dealers amplify them.
Real-time delta of market-orders vs resting liquidity. Persistent imbalance flags one-sided flow before it shows on the chart.
Heavy aggressive volume printing into a level with no follow-through - large hidden liquidity is soaking it. Reversal candidate flagged.
Stop-run beyond a key level followed by an immediate reclaim. Classic liquidity-grab pattern, auto-detected and timestamped.
Price making new extremes while CVD fails to confirm. Underlying flow disagrees with price - exhaustion signal.
Seven cross-validated readings of realized volatility, plus implied-side fits, all reconciled into one volatility regime call you can trust.
If you want to audit the math before you trust the output: here it is. The full quant matrix and a long-form catalogue of every research-grounded indicator with paper, year, and authors.
57 methods across 9 domains. Each is a real implementation tied to a real paper. Hover any cell for the citation.
Hover any cell to see the citation. Every one is a live function in the codebase, not a marketing checkbox. Long-form descriptions of the headline 21 →
CharmCast generates a 250-500 word AI briefing every market open that reads like the desk notes institutional traders pay $30k/year for. Cross-asset chains of causation, calibrated language, three-horizon framing (tactical / cyclical / structural), and explicit trip-wires — not generic "stocks went up" pap.
Every claim is built as a chain: "oil down 12% → eases inflation expectations → 10y holds 4.34% rather than ripping → supportive for long-duration tech." No isolated factoids.
Tactical (today) / Cyclical (1-3 months) / Structural (6-12 months) explicitly labeled. Stops the common reader mistake of treating intraday signals as structural.
No "screams" or "classic" — every causal claim is paired with the alternative explanations and what would distinguish them. VIX up on a green day gets three competing interpretations, not one.
Global Liquidity Pulse · Recession Ensemble · Policy Path Gap · Credit Stress Radar · Yield Curve Regime · Labor Market Breadth. Each updated continuously with status pills + sparklines.
The macro read isn't isolated — it feeds straight into the symbol-level analysis prompt as the "MACRO ANCHOR" so your trade plan's bias is grounded in the cross-asset regime, not just the chart.
tape modestly firmer on the surface but extremely narrow underneath. spx +0.84%, sector breadth 5 green / 6 red — the headline number is a momentum-chasing tech story rather than broad risk-on. ndx +2.35% doing all the heavy lifting; rsp (equal-weight) -0.10%, so the "S&P 493" is essentially flat.
energy is the swing factor and it's against us today: brent +1.23% at $101.29, wti +0.64% at $95.42. crude above $100 keeps inflation expectations sticky and removes the narrative support for fed cuts → 10y holding 4.34% rather than fading toward 4.20%, which is exactly the regime that compresses long-duration multiples. if oil rolls back through $100 the whole tape re-rates higher; if it pushes toward $105 yields back up and tech gives back the day's gains fast.
dxy weak (-0.2%), gold firm, btc unchanged — that mix usually says "macro is fine, liquidity is thinning at the margin." europe lagging US peers (DAX -0.4%) is the dollar-down + crude-up combo biting in EUR specifically. dealers short-gamma at -1.4B GEX with flip distance 0.34% — today's tech rip is partly mechanical, expect a snap-back if the AI-capex catalyst fades into the close.
cyclical (1-3mo): liquidity is draining at the short end. crude holding $100+ keeps inflation sticky and removes the Fed-cut narrative. size INTO concentration, don't chase extremes.
structural (6-12mo): NFCI -0.51σ (easy by ~0.5σ). global liquidity sitting at TROUGH which historically supports risk at 9-12mo lag. recession gauges 1/8 firing — no broad-based recession signal.
CharmCast doesn't just throw black-box "signals" at you. Every layer of the stack - vol surface, microstructure, macro nowcast - is composed from peer-reviewed methods. Here's the academic backing for what runs every minute under your tape.
Per-contract Lee-Ready classifier: quote rule, tick test, prior-tick zero-tick fallback. Aggregates 15-minute net signed call/put flow × delta into a live dealer-net-delta read - fresh-flow positioning the static OI-based GEX can't see. Distinct from put/call ratio: it's signed and contract-by-contract.
Three Hawkes (1971) processes - trades, big prints, sweep clusters - fitted online via exponential-kernel MLE. The branching ratio n = α/β is the gauge: ≥ 0.75 critical, 0.55–0.75 elevated, < 0.30 quiet. Bacry-Mastromatteo-Muzy showed n leads realized-vol shocks; we surface it composite-style with the active driver tagged.
Five-parameter Stochastic Vol Inspired fit per expiry - denoises wing IVs, surfaces butterfly + calendar arbitrage flags in closed form. Levenberg-Marquardt damped Gauss-Newton solver under the hood.
R/S analysis of realized-vol log-paths reveals the persistent anti-correlation Gatheral-Jaisson-Rosenbaum first showed in 2014. H < 0.5 signals mean-reverting vol regime - a tradable edge before VIX confirms.
Per-strike charm × time-to-close projected forward in 30-min buckets gives a cumulative dollar-delta hedge demand. When projected hedge magnitude exceeds 1× ADV, we flag a forced-flow regime.
Carr-Wu's formal VRP from the Demeterfi-Derman-Kamal-Zou variance swap strip vs the live VIX-future quote. Persistent positive basis = realized-vol risk premium being paid; negative = market expecting downside vol surprise.
Dollar-vanna across the chain × 1pt VIX move = dealer hedge demand around vol triggers. Barbon-Buraschi documented the pre-FOMC vanna-driven pump that flips on the announcement; we surface live size-of-flow.
Cont-Kukanov-Stoikov's OFI extended through 5 book levels with exponential depth weights w_k = exp(−κ·k). Dominates top-of-book OFI for short-horizon mid-price prediction. Spoof-discounted via our 3-feature classifier.
Stoikov's volume-adjusted micro-price (depth-weighted across 5 levels) minus the standard top-only microprice gives a directional drift signal in the next-tick move - quietly leading mid by ~50ms on average.
Cont-Larrard's Markovian LOB model: P(next mid up) ≈ (λ_bidArrive + λ_askCancel) / Σλ. We track all four EW-averaged intensities at 5-second time constant. Flips bullish/bearish before the mid moves.
Cancel-to-trade ratio over rolling 30s + median milliseconds-to-recover-50% post-sweep. Egginton-Van Ness on CTR as quote-stuffing flag, Large on resilience half-life as a regime-shift early-warning.
Square-root-law impact (Almgren-Chriss style) with a 1000-sim Box-Muller bootstrap p-value on the residual. We label every flow-classified run with HIGH/MEDIUM/LOW confidence - no more vibe-based "absorbed" calls.
Z-scored HY (BAMLH0A0HYM2) + IG (BAMLC0A0CM) OAS against rolling 1y baseline → composite stress signal. BIS Working Paper 1272 (2025) shows credit-spread regimes lead equity vol with 5-15 day lag.
Geopolitical Risk index shock z > threshold triggers a sector-by-sector forward-vol-bump forecast using empirically calibrated betas: Energy +1.40, Defense +1.20, EM +1.10, Tech −0.60, etc.
Bańbura-Modugno-style composite from 6 high-frequency proxies: copper/gold ratio · 5y5y forward inflation · HY OAS · WTI · baltic dry · retail tracker. Bucketed EXPANSION → CONTRACTION; Δ vs prior is the trade-relevant read.
Daily returns of MTUM · QUAL · USMV · IWD · SIZE · COWZ ETFs, ranked. Leader/laggard pattern maps to canonical regimes: momentum led = risk-on, USMV-led = late-cycle defensive, IWD-led = value rotation away from growth.
Heston stochastic-vol pricer extended with Merton-style log-normal jumps via Poisson thinning. Three jump params {λ, μ_J, σ_J} fix the systematic under-pricing of short-dated wings the pure SV model misses.
For every watchlist symbol: ATM-IV, 25Δ risk-reversal, IV-RV spread → z-scored within sector. |z| > 1.5 with rising OI flags single-name catalysts before the headline hits the wire.
Short-horizon predictor based on call-put implied-vol differential. Zhou-Wang-Lai-Yen 2024 showed the rolling 5-minute Δ leads SPX 30-min returns; positive Δ-DIF = call-side IV firming first.
VIX9D / VIX / VIX3M slope + ratio9d → 4-state regime (BACKWARDATION · STEEP CONTANGO · DEEP CONTANGO · INVERTED). Front-end stress flag fires when VIX9D > VIX by ≥ 1pt - reliably leads vol-shock days.
Yang-Zhang · Parkinson · Garman-Klass · EWMA RiskMetrics · robust median + Bollinger and Keltner bands. Hurst + Variance Ratio + Choppiness Index decide which estimator the AI weights for the regime call.
Every reading flows into the same prompt the AI builds your trade plan from - and into the knowledge graph so you can trace which research-grounded indicator drove which call. See the tour →
If you've already trialled GEX Radar, SpotGamma, or CoreConvexity — here's exactly where CharmCast sits in the same room, capability for capability, dollar for dollar.
Most tools stop at gamma walls and a daily PDF. We are an end-to-end terminal — read structure, compose a plan with the AI, route the bracket, and let the journal auto-grade you. Side-by-side against the three platforms most traders compare us to.
Six patterns the existing gamma platforms don't do at all - and the structural reason CharmCast can.
GEX Radar shows you walls. SpotGamma sends a daily PDF. CoreConvexity shows dealer positioning. Each is a single layer. CharmCast feeds structure + macro + microstructure + orderflow + ICT + vol framework + recent news into one prompt before the AI writes your call. The model sees the whole picture every time.
SpotGamma's daily report is a static document. CharmCast's analysis is a starting point - the AI Chat tab lets you ask follow-up questions ("what invalidates the long?", "if gamma flip breaks, what should I expect?") and the AI sees the same metrics + your saved research notes when it answers. Refresh + Check pulls fresh data and compares to the prior call without burning a full re-analysis.
Drop your research notes, macro thesis, or trade-style preferences into the Research vault. Every future analysis and briefing is generated with your context injected as authoritative. A swing trader and a 0DTE scalper get fundamentally different outputs from the same market state.
The closest competitor has alerts. CharmCast generates a strict-JSON plan — entry, stop, multi-tier targets, breakeven trigger, trailing rules, time stop, news kill switch, conviction grade A+ to C. Review it on the Executions tab, then SEND BRACKET routes through your connected broker (Tradovate, IBKR, Alpaca, NinjaTrader, MT5, cTrader, plus 13 prop firms). Server-side risk gates verify before any order fires.
Most gamma platforms ignore microstructure entirely. CharmCast ingests L2, OFI (order flow imbalance), cancel rates, VPIN, footprint, aggressor delta, and ICT patterns alongside the gamma surface. The AI prompt explicitly says "dealer regime + tape + cancel rate + footprint headline" - so when dealers are short γ AND VPIN turns toxic, you see the combined signal flagged, not two charts in two tabs.
Every trade you send through CharmCast - manual entry, AI plan execution, or CSV import from your prop firm - lands in the journal auto-graded A+ to F by R-multiple, with the AI's original reasoning preserved. Statistics tab shows your win rate by setup grade, per-symbol P&L, equity curve. The competitors give you levels; CharmCast tells you which of your trades on those levels actually paid.
Most "AI trading" tools ship overfit models with no honest evaluation. CharmCast's forecasting layer is built like a research lab — every model passes the same statistical bar professional quant desks use before it ships to your dashboard.
Every volatility forecast comes with a calibrated lower / upper range — the model tells you the band the next move is likely to land in, with a real coverage guarantee. You see uncertainty as a number, not a hand-wave.
The forecasting layer ingests the live order-flow, dealer Greeks, microstructure, and macro context that already drives the dashboard. No mystery features — if you can see it on a chart, the model saw it during training.
Walk-forward backtests, purged k-fold cross-validation, Diebold-Mariano significance tests, Model Confidence Set membership — the same discipline academic vol-forecasting papers use. A model only ships if it beats the standard baseline on data it has never seen.
Every feature is traceable to a published paper — Corsi's HAR-RV, Cont-Kukanov-Stoikov order-flow imbalance, Chevyrev-Kormilitzin path signatures, Andersen-Bollerslev jump separation, López de Prado's purged validation. If you want to audit the math behind a forecast, the citation list is in the dashboard's research tab.
We will never tell you we can predict tomorrow's SPY direction. Nobody can. What the forecasting layer does is quantify regimes, calibrate vol envelopes, and flag toxic-flow conditions early. The job of converting that into a trade is still yours — informed by better data, not replaced by it.
The pipeline keeps a rolling window of historical L2 depth and trade data, deep enough that the model sees multiple volatility regimes — not just last week. Rebuilds nightly. Older signals don't drift out of relevance just because the market changed last quarter.
20+ broker integrations including 13 prop firms. Three subscription tiers. Cancel from the billing portal anytime; your journal and saved research stay on the account.
Direct retail and prop firms - bracket orders, position streams, fill events, all with risk gates enforced before send.
Side-by-side against the platforms most traders evaluate alongside us — at the most expensive tier each one publicly sells. CharmCast STARTER stays $19/mo and still covers the surface.
Three tiers. Starter is the read-only data stack at $19. Pro adds AI in the loop, broker routing, and the new orderflow terminal early-access at $49. Max unlocks Opus on every analysis, the full live terminal, custom anomaly tuning, and ML signal previews at $99. Same prices forever for early adopters.
For traders who already know how to read the data.
Live trading. AI in the loop. Custom dev.
For traders who run real size.
No. CharmCast uses server-side keys for all AI calls. You're rate-limited by your subscription tier instead of paying per-token.
Never. The execution planner produces a plan. You review every field on the Executions tab and explicitly click SEND BRACKET. Risk gates run server-side before any order fires, and dry-run mode is the default until you flip it off.
Yes - Apex, TopStep, Earn2Trade, Bulenox, MFFU, MyFunded, FTMO, FundingPips, FundedNext, FXIFY, The5ers, E8 Funding, Goat Funded Trader. Picking the prop firm in the connect dialog routes through the underlying broker (Tradovate / ProjectX / MT5 / Match-Trader).
Paper mode mocks every order so you can test the workflow without risking capital. Live mode sends real orders through the broker. Live execution is gated to Pro and Max tiers and requires explicit per-session confirmation.
The Predictions tab tracks every level call the AI made and marks them HIT / MISS / EXPIRED at end-of-day. Per-AI accuracy stats build over time on your dashboard.
Yes - drop a CSV from Apex, TopStep, MyFunded, FTMO, Tradovate, or NinjaTrader. The format is auto-detected and every row is auto-graded A+ to F by R-multiple, then folded into your equity curve and per-symbol stats.
The Terminal — full footprint, 10-level DOM ladder, bookmap-style liquidity heatmap, AI signal sidebar — is in private build right now. Pro tier gets early-access betas as panels stabilize; Max tier gets the full feed at launch with no waitlist. Free / Starter accounts see a "coming soon" placeholder with a Discord link for ship-day updates.
250-500 words written in lowercase narrative voice, generated every market open. Three-horizon framing (tactical / cyclical / structural), explicit chains of causation ("oil down → eases inflation → 10y holds → supportive for long-duration tech"), every causal claim paired with the alternative explanations. Plus a structural-read paragraph covering global liquidity / NFCI / recession gauges, and 3-6 corporate-news bullets. Pro and Max tiers; delivered in-app and to your inbox.
Starter sees the raw composite anomaly list (detector + confidence + timestamp). Pro adds AI commentary explaining why each one fired and what to act on. Max adds custom threshold tuning — you pick which detectors fire, adjust z-score cut-offs, and compose your own signal stack.
Cancel anytime from the billing portal. Your tier downgrades at the end of the current period. Trade journal and saved research stay on your account regardless of tier.
Institutional-grade options intelligence. AI-routed trade plans, server-side data pipeline, your broker.