Paper first · Stocks & crypto · Fail-closed guardrails

Deploy trading bots that trade your rules while you are offline

Compose, backtest, and run algorithmic bots across stocks, ETFs, and crypto — with fail-closed pre-trade guardrails on every order.

No credit card · Paper trading is free · Live execution is gated and off by default

Offlinr

Offline? Offlinr's online.

Trading bots that run your portfolio while you are offline — event-sourced, streaming, paper-first.

Four steps from idea to guarded execution

The same decision path powers backtest, paper, and live — so the logic you validated is exactly the logic that runs.

01

Discover

Scored smart lists rank a screened universe by recency-decayed signal composites, so the freshest setups float to the top — not stale ones.

02

Validate

Walk a fitted model forward over up to ten years of daily bars with embargoed splits and champion/challenger lineage — no peeking, and the bias is declared.

03

Deploy

Launch on the built-in $100k paper account in minutes, with every order through a buying-power, kill-switch and Reg-T margin gate.

04

Relax

Optional drawdown guards, stale-data fail-closed gates, and a full audit trail watch every tick — so a bot is never running unwatched.

You can see exactly how a bot decides

No black box, and nothing you have to take on faith. Every order traces back through the strategy that placed it, the composite that triggered it, and the individual algo scores that made up that composite — the same picture you get while you are still configuring the bot.

Universe~1,400 symbols, rescreened live32 signal algoseach scores every symbol, −1 … +1Blend + gatesweighted average, gates multiply itStrategy → ordersrisk-gated, event-sourced
How one decision is assembled, each layer made of the one beneath it. Schematic. Counts are the platform's real shape; no returns are shown or implied.

Every algo shows its own score

32 signal algos, each producing a score between −1 and +1 with its own confidence. You choose which ones run and how much each one counts.

Gates multiply, they don't vote

A regime gate has no direction of its own — it scales everything else. A gate reading 0.2 quietly divides every other algo's influence by five, which is why the builder draws it rather than listing it.

The strategy is the only thing that trades

One composite per symbol goes to the strategy, which turns it into orders under your thresholds — then risk checks, buying-power pre-checks and the kill switch all get a veto.

And it is all replayable

Orders, fills, cash and positions are projections over an append-only event log, so the reason for any decision is still there weeks later.

A real strategy library, built in

Every card is a real strategy from the platform catalog, and every diagram is the same one the app draws from its registry — the actual mechanic, redrawn from the parameters shown. Illustrative geometry, never real returns.

DCA — Martingale

AccumulationStocksCrypto

Buy a base order, then bigger 'safety' buys on each dip; sell the lot at a small profit.

Backtest it in the Lab

Supernova

Volatility harvestCryptoStocks

Scatter a small seed across many top movers, then harvest winners and recycle the cash into martingale accumulation.

Backtest it in the Lab

Cross-Sectional Momentum

RotationStocks

Rank the whole universe and hold the top of it — optionally shorting the bottom.

Backtest it in the Lab

Mean-Reversion Fade

ReversionStocksCrypto

Buy the oversold extreme, sell into the recovery — the contrarian of the set.

Backtest it in the Lab

Range Grid

RangeCryptoStocks

Rest a ladder of buy rungs across a sideways band and sell the stack at one target.

Backtest it in the Lab

Bracket (TP/SL)

Risk-firstStocksCrypto

One signal entry, then the exit is price — a take-profit above and a stop below.

Backtest it in the Lab

The platform under the bots

Retail-grade UX on top, institutional-grade machinery underneath.

Signal algorithms
32
built in
Execution strategies
27
composable
Market data
Live marks
streamed, REST gap-filled
Money path
Event-sourced
append-only log
From the appCapability counts from the registries — never users, assets or performance

Scored smart lists & signals

Live screens that rank symbols by a decaying composite of the algos you pick — up to 8 of the 32 built in. Bots subscribe to the list, not a static basket.

Walk-forward Lab

Calendar walk-forward backtests with embargoed splits, out-of-sample scoring, and explicit champion promotion — models earn their place.

Broker-agnostic

Alpaca for stocks and crypto today, behind a broker-adapter contract that adds a venue without touching a strategy.

Risk guards & drawdown ladders

Optional drawdown ladders, buying-power pre-checks, kill switches, and pre-trade gates on every order path. Missing data means no live trade.

Real-time observability

OpenTelemetry traces, structured logs, and live health dashboards across every money path — the platform watches itself, in the open.

Multi-tenant, audit-grade security

Tenant-scoped RBAC, encrypted broker credentials that are never returned by the API, and a hash-chained audit row on every mutating request.

The product, not a picture of it

Both panels below are the application's own components, running on this page: the blueprint the bot builder redraws as you configure, and the plain-English account every dry run carries. The inputs are examples and say so — nothing here is a price, a position or a return.

The desk is behind the sign-in, on purpose

Positions, working orders, guard telemetry and the live tape share one pane — and every one of them reads your own account, so there is nothing truthful to show a stranger. It is the dashboard you land on when you sign in.

Open it on paper

Built for trust, down to the ledger

Everywhere money moves, the platform is engineered like a ledger, not a spreadsheet.

  • Event-sourced money paths

    Positions, cash, and P&L are projections folded over an append-only event log — never a mutable balance that can silently drift.

  • Fail-closed by default

    Unknown flag: off. Missing or stale data: no live trade. Reconciliation mismatch: halt and alarm. The safe state is always the default state.

  • Hash-chained audit log

    Every mutating request — accounts, orders, bots, risk settings — appends a hash-chained audit row: a semantic record where the action has one, the route and principal otherwise. Tamper-evident by construction, re-verified nightly, anchored offsite.

  • Encrypted credentials

    Broker keys are encrypted at rest and never logged or returned by any API response.

audit_events · hash-chainedschematic
  1. #104order.submittedSYMA · buy 12hash 9f2c
  2. #105order.filled12 @ $50.25prev 9f2c → hash 41be
  3. #106ledger.reconciledcash −$603.00prev 41be → hash c803
projection ▸positions · cash · P&L= fold(events #1…#106)

Balances are derived from the chain and can be re-folded from it. Example rows, not real events.

Put a disciplined bot to work

Start on the built-in $100k paper account — no card, no broker required. Live execution is a separate, gated step.