How HelmFi works

1. Watch the live demo (no signup) to see the product running on a real paper account.

2. Join the waitlist — HelmFi is in private beta; access opens in cohorts.

3. Once invited: sign in, connect an exchange with a trade-only API key (or no key at all for paper mode), and pick the Helm Index.

4. Your account starts in paper mode. Watch it for as long as you like — there is no pressure and no time limit.

5. Going live is an explicit opt-in: Pro tier, stored trade-only keys, and a written risk acknowledgment.

HelmFi never holds your funds. You create an API key on your exchange with TRADE permission only — withdrawal permission stays disabled on the exchange side, which means the key physically cannot move funds out even if it leaked.

Keys are encrypted at rest with a server-side master secret and are never shown again after saving. You can revoke the key from your exchange at any moment, which instantly severs HelmFi's access.

Paper mode requires no keys at all — the bot simulates fills against live market prices.

Paper mode runs the identical engine — same signals, same regime gate, same risk stops — against a simulated balance with realistic fees and slippage applied. It exists so you can verify behavior with zero risk.

Live mode places real orders on your exchange account. It requires the Pro tier, stored trade-only keys, and an explicit risk acknowledgment. You can stop the bot at any time; stopping halts trading but never sells your holdings.

Our public demo page is itself a paper account running 24/7 — that is exactly what your paper account looks like.

A basket of ~11 major cryptocurrencies, equal-weighted, volatility-targeted, and gated by a walk-forward-validated trend-following regime rule: when the trend turns down across the universe, the index steps toward cash instead of riding the drawdown. (A separate AI classifier provides live market-regime context on your dashboard and powers chat.)

It rebalances on a 72-hour cadence, selling winners back to target weight (structural profit-taking). Drawdown control comes from the regime gate plus vol-targeting — a per-position stop was tested and removed 2026-06-13 after a validator A/B found it return-negative.

In the 2021–2026 validation it returned +411% vs BTC's +133%, with a worst drawdown of −25.8% vs BTC's −75%. These are historical validation artifacts, not return promises — the full evidence package, including its limitations, is on the Helm Index page.

Every strategy must pass: a bootstrap 95% Sharpe confidence interval above zero, permutation alpha p < 0.10, walk-forward windows overwhelmingly positive, and bear-market drawdown contained relative to benchmark.

Eighteen of our own strategies failed this gate. They are published in the strategy graveyard with the reason each one died — overfitting, fee drag, statistical noise, regime shift.

You can run your own backtest through the same gate on the validator page: upload a freqtrade result or an equity CSV and get the same verdict our internal strategies get.

Account stop-loss: set your own halt threshold (5–50% below your account's peak), with optional wick protection that requires the breach to persist before firing. Checked every 5 minutes against your real exchange balance.

Regime gate + vol-targeting: the index's drawdown control. A per-position stop was removed 2026-06-13 — a validator A/B found it return-negative and drawdown-neutral (it cascade-liquidated the basket on routine dips).

Drift freeze: if your exchange holdings ever diverge from what the bot expects (manual trades, partial fills), it halts rather than trading on stale assumptions.

Daily-loss kill switch: tier-dependent daily realized-loss limits that stop the bot and alert us.