[ Auracle / v1.0 ]
Backtest, schedule, deploy strategies — on hardware you control. AI-native via MCP for Claude / Cursor / Codex. Bring your own broker.
What QuantConnect would be if it ran on your laptop.
One-line install. Pre-built Docker images. ~3 minutes from curl to running.
[ Built on tools quants already know ]
[ The problem ]
Every signal, every backtest, every fill — visible to the vendor. Their outage is your downtime.
Per-datapoint, per-minute, per-algorithm fees. Heavy iteration costs more — structurally hostile to actually doing research.
Account suspended mid-trade or after a ToS shift, you lose more than convenience — you lose your operations.
Cloud platforms backtest in one environment and trade in another. Live divergence is the rule, not the exception.
[ The Auracle approach ]
Self-hosted Docker stack — Mac, Linux, or VPS. Web UI, scheduler, JupyterLab,
MCP server, all sharing a local TimescaleDB. The same
Strategy class powers your backtest and your
live cron. No execution gap.
[ 01 ]
Backtesting
Pandas-native, with transaction costs, slippage, sizing, and a proper 1-period
lag. Walk-forward built in — 2-year train, 6-month test, rolled across
history. The same Strategy class runs the
backtest and the live cron. No research-to-production gap.
[ 02 ]
Broker bridge
Auracle plugs into the broker accounts you already have at IBKR,
Alpaca, ClearStreet, and Hyperliquid — using each broker’s
native SDK, with your credentials stored only in your local .env.
We don’t take custody of your funds, don’t see your fills,
don’t sit between you and the broker. Auracle is software; the
broker relationship is yours.
ib_insync (native macOS IB Gateway or Dockerized)alpaca-py SDK)[ 03 ]
AI-native
Built-in MCP server wires Claude Desktop, Cursor, or Continue straight to your install. Ask Claude to run a backtest, ingest a year of SPY bars, or audit recent fills. 23 tools, bearer-token authenticated, scoped to your install. No other self-hosted platform ships this.
[ 04 ]
Research
JupyterLab ships with auracle.* pre-installed and
the launcher tile pre-configured. Open a notebook, import
fetch_daily_bars_safe, and you’re already
wired to your broker. No setup, no env vars to plumb.
auracle.* pre-installed and tested[ Where your data lives ]
Only outbound traffic is the paid-tier license heartbeat every 10 minutes — install UUID, key, version. Strategies, fills, P&L never leave your machine. Community runs fully offline.
[ By the numbers ]
Comparison reflects public-doc positioning of major cloud quant platforms, May 2026. Verify with their pricing pages before deciding.
[ Install ]
Multi-arch images (amd64 + arm64) on Docker Hub. The bootstrap generates secrets,
pulls the compose file + images, and brings up the stack. ~3 minutes,
curl to running.
$ curl -fsSL https://raw.githubusercontent.com/SiixQuant/auracle-installer/main/install.sh | bash
Requires Docker Desktop or Docker Engine. macOS, Linux, and any Docker-capable VPS. Apple Silicon native.
Read full install docs ↗[ Pricing ]
No usage metering. No per-data-point fees. No surprise bills.
[ Tier 1 ]
Perpetual community license. For researchers, hobbyists, and anyone evaluating before committing.
[ Tier 2 ]
For quants running production research. Try Community first — upgrade in 30 seconds when you outgrow it.
Keys bind to a single install at activation time — server-side enforced. Refunds revoke within seconds when the billing webhook is configured; within 10 minutes otherwise. Bulk pricing for >5 installs: contact@aurapointcapital.com.
[ FAQ ]
The runtime engine ships as pre-built Docker images; source is private. Installer, client SDK, docs, and EULA are public. Open enough to audit what you’re running, closed where it would expose the licensing system.
1-hour grace (full function), then 24h degraded (read-only, no new schedules or live orders), then BLOCKED (read-only). Refunds trigger revocation within seconds when the webhook is configured; within 10 minutes otherwise. Your strategies, notebooks, and history stay on your hardware — Auracle never touches them.
Yes — that’s the design point. Execution uses your broker’s native SDK (ib_insync, alpaca-py, hyperliquid-python-sdk); we don’t sit between you and the broker. Risk gates (sector caps, exposure limits, NLV-aware sizing) are configurable.
Most quant platforms are cloud-first and usage-metered — you pay per backtest minute, per data point, per algorithm. Strategies and data live on their servers. If you want both to stay on your own hardware, Auracle is the answer.
Auracle bridges to your existing broker accounts at IBKR, Alpaca, ClearStreet, and Hyperliquid (Tradier + Tradovate planned for v1.1) using each broker’s native SDK. 23 MCP tools let Claude / Cursor / Codex drive the platform via the agentic stack; walk-forward backtesting + ML training are built-in; pricing is flat $199/seat/mo. No metering, no per-data-point fees, no middleman holding your fills.
You need to write a class with a prices_to_signals(prices) method that returns a DataFrame. That’s the minimum. The bundled example is ~30 lines; most user strategies are 50–200. Comfortable with pandas? You’re fine.
No. Keys bind to a single install at activation time — the second machine fails server-side. Multiple installs need multiple licenses. Email us for >5 installs.
Nowhere. Bars, strategies, orders, fills, P&L all live in the local Postgres on your hardware. Only outbound call is the 10-minute license check to our license server (install UUID, key, hostname, version, timestamp). EULA section 4.
Run auracle license activate <your-key>, or paste at /ui/license. Effective on the next minute-tick. No reinstall, no data migration, no downtime.
Install Community in three minutes. Upgrade when you’re ready.