Your AI trades it. Coil teaches it how.

Coil is a rules-based, long-only stock scanner and trading engine that finds market leaders at real pullbacks — and can execute those same rules through your own brokerage account, run by your own AI agent. It ships as three parts in one download, and you run it yourself. The scanner scores every name in the Equities book (S&P 500, Nasdaq-100, and a Macro book of bonds, income, metals, commodities — ~560 names) — opportunity, entry-window, hold-conviction, leadership, sector-rotation phase, market posture — and maps leaders to leveraged vehicles (NVDA → NVDL, QQQ → TQQQ). The dashboard is the command view. The engine trades the scanner's published signals: it buys the highest-conviction leaders when they're ready, sizes by conviction, accelerates with leveraged ETFs, exits on a ladder with trailing stops, and holds cash when the pool is thin — all behind an immutable safety core. It runs on your machine, with your agent, your broker, and your keys.

It is built for Claude Code, and any capable coding agent works — Cursor, Codex CLI, Gemini CLI, or anything that can read files, run Python on a schedule, and speak MCP. The engine plans and prints the orders; your own agent places them through a broker connector — built for Robinhood, and any broker with an equivalent MCP works. Under the hood it's Python on Mac or Windows (the agent-led onboarding detects your OS) with a free Alpaca data key powering the market-wide scan — trading and live intraday data use the broker connection itself. You add your own broker account id and key, and your credentials never leave your machine.

Once armed, it runs hands-off: it scans the market, takes the entries, manages every exit by rule, and applies risk circuit-breakers tied to your account's high-water mark — no screen-watching required. You're never locked out: read every rule, see exactly what it did and why on the local dashboard, and stop it in one command. The circuit-breakers fire by rule, without emotion — they reduce big single-day damage, they don't remove risk. It uses leveraged ETFs to accelerate its best ideas, and those can move fast and lose value rapidly, including total loss.

All three parts in one download: the market-wide scanner (opportunity, entry-window, hold-conviction, leadership, sector rotation, market posture, with leveraged-vehicle mapping), the command dashboard (books, scorecards with why-this-score notes, top picks, structural stops and targets), and the long-only autonomous engine (conviction-ranked entries, leveraged acceleration, ladder exits, trailing stops, cash when thin). Plus the immutable safety core with persistent circuit-breakers, the local sizing-only learning loop, the validation harness with a self-test suite of 100+ assertions you can run yourself, and the plain-English onboarding. No credentials are included — you add your own.

Because leadership rotates. Every cycle has leaders — semiconductors led in 2024, but before and after that the baton passed to other groups. A one-sector engine is right only while its sector leads and dead weight the rest of the time. Coil scores the whole market — the Equities book of S&P 500, Nasdaq-100, and Macro names — and follows leadership wherever it goes, buying names coiled at real support when they're ready and standing aside from the falling knives and the overbought chasers. The name still fits: Coil looks for leaders coiled at support and strikes when they're ready; it just does it across the whole market now.

Very. Coil accelerates its highest-conviction ideas with 2× and 3× leveraged ETFs, so a ~10% move in the underlying is roughly a ~20–30% move in the vehicle, before gaps and slippage. Leveraged ETFs also decay on multi-day holds because they track daily moves, and a position can gap through the hard stop before the open (a stop can't fill in a gap). The engine caps exposure and the circuit-breaker ladder reduces single-day damage, but neither removes risk. You can lose money rapidly, including total loss.

The engine only buys when the pool has real leaders that are ready — coiled at support, not falling and not overbought. When the market posture turns defensive it tightens entries and rotates toward the Macro book (bonds, income, metals); when nothing qualifies it holds cash, and on Robinhood Gold that idle cash earns the broker's cash sweep, ~3.35% APY (variable) as of Sept 2026 — the broker's yield, not Coil's, and not guaranteed or risk-free. The research backbone is honest about when the edge shows up: through end-2025 the ranking brain ran about even with SPY at roughly a third less drawdown, and it pulled ahead in leadership regimes (2025 and 2026). Those are research figures, not a prediction.

The engine, no — the $79 download is a long-only equities engine and does not trade crypto. The board, yes as a read: since July 15, 2026 the Coil Scanner board pairs the Equities book with the Crypto book — BTC and ETH scored with long-only trend signals. The honesty rider ships with it, ugly parts first: the cold backtest does not beat buy-and-hold CAGR (59.6% vs 74.9% for a 50/50 BTC/ETH hold, 2016–2026); what it buys is smaller crashes (2022: 0.0% vs −66.0%; max drawdown −64.3% vs −87.4%) at the cost of whipsaw years (2024: −22.6% vs +74.7%). Signals tracked live since July 15 — backtest before that, not traded with real money. The full numbers, losses included.

You buy the current version once and keep it. New validated versions — new scanner factors, entry windows, exits, or book coverage — are cold-validated first and announced on coil.trade; there's no auto-update service and no subscription. What you download runs forever, offline, on your own machine, and its local loop keeps tuning position sizing to your own fills within hard bounds.

Running it is what Coil is built for: an agent can discover it (there's a machine-readable manifest at /.well-known/coil.json), install it, and operate it on a schedule. Buying is ordinary web commerce: checkout is a hosted Gumroad flow — Gumroad is the merchant of record and handles receipts and license keys. It's a single one-time purchase, and all sales are final (instant digital download). There's no special agent-checkout integration; an agent with delegated card authority can complete the hosted checkout like any web checkout. And the scanner board itself is agent-purchasable natively: $0.10 per read in USDC over the x402 protocol — no account, no card, no email (the agent endpoint). It runs on your approval and your own credentials, and you stay in control. To be clear: this is software you run yourself. It's not investment advice and makes no income promise.

About an hour to a scheduled dry-run. You point your agent at the folder and the interactive onboarding takes over: it detects your OS (Mac or Windows), sets up a free Alpaca data key, and runs first light — the agent introduces itself, scans the market, and opens the dashboard with live scores within minutes. Then it watches in dry-run. Going live is a separate, deliberately gated flow: your own account allowlisted, an integrity re-pin, the self-test suite green, and a typed total-loss acknowledgment. It ships disarmed.

The ranking brain — the leadership scoring the whole system is built on — was replayed on a survivorship-free research backbone: point-in-time S&P 500 membership including delisted names (so failures like FRC and SIVB are in the data, not quietly dropped), next-open fills, and modeled trading costs, from 2017 through 2026 H1. That's research that validates the brain, not a live account record, and it has self-tuning degrees of freedom, so treat the figures as evidence, not proof. The live engine is newly running; live and forward results, accumulated on small size over time, matter more than any backtest — and every figure is re-runnable yourself with the included harness.

Coil trades actively — many entries and exits a year across a rotating set of names — so realized gains are largely short-term (taxed as ordinary income), and active trading can create wash-sale complications that defer losses. None of that is tax advice — talk to a tax professional about your own situation before running it with real money. Coil is software you operate yourself and reports nothing on your behalf.

No. Coil is software you run yourself, with your own credentials and your own capital. It is not investment advice, not a managed fund, and not a guarantee of profit. Coil holds no funds and executes nothing on anyone's behalf. Trading leveraged ETFs carries substantial risk of loss, including total loss. The performance figures on this page are a survivorship-free research backtest that validates the ranking brain — not a record of anyone's client or account returns, and not a prediction.