SAEBL
AUTONOMOUS TRADING AGENT

Other bots
follow rules.
This one adapts.

Saebl analyzes markets, explains every decision in plain language, and executes under mandatory risk limits. Every trade is logged, every decision is auditable, and every lesson is tracked. Autonomous trading with guardrails you can verify.

US equities via Alpaca · Intraday · Commission-free · Interactive Brokers coming soon

See How It Works
saebl — decision trace

regime: BULL_CALM  confidence: 0.82

analyzing NVDA...

→ Semiconductor sector showing relative strength

→ Pullback to 50-day MA in uptrend

→ No existing position in sector

→ Past lesson: momentum entries in calm

  regimes have 68% win rate

decision: BUY  qty: 15  confidence: 0.78

stop_loss: $173.20  take_profit: $186.50

risk: 0.8R  bracket_verified: true

THE PROBLEM WITH EVERY TRADING BOT

Black box decisions

You never know WHY it bought. The output is a trade log — entry, exit, P&L. That's it. No reasoning. No context. No explanation you could learn from.

Zero learning

Every platform is stateless between trades. Lost money on momentum plays in choppy markets? It'll do it again tomorrow. And the day after. Forever.

Regime blind

Bull market or crash — your bot runs the same strategy. Professional traders adapt to conditions. Every retail bot pretends conditions don't change.

Risk is a settings panel

"Set your stop-loss percentage." That's the entire risk system. No daily limits. No portfolio awareness. No auto-kill when things go wrong. One bad day can erase a month.

HOW SAEBL IS DIFFERENT

It explains every decision

Every buy, every hold, every sell comes with a plain-language explanation. Not a signal number. Not an indicator reading. A full chain of reasoning you can read, understand, and learn from.

reasoning:

  "NVDA pulled back to 50-day MA

   while sector shows relative

   strength. Historical win rate

   for this pattern in BULL_CALM:

   68% over 23 trades."

confidence: 0.82  →  auto-execute

It learns from every outcome

After every trade closes, Saebl analyzes what happened, extracts a lesson, and updates its playbook. Lessons gain evidence over time and graduate from hypothesis to validated rule.

lesson: "momentum_calm_regime"

status: VALIDATED

  evidence: 23 trades

  win_rate: 68%

  avg_r: +1.4R

rule: "In BULL_CALM, momentum

  entries on MA pullbacks with

  sector strength are high-

  probability setups."

It reads the room

Saebl detects market regime from SPY, QQQ, DIA, and VIX. Bull or bear. Calm or volatile. It adjusts position sizing, strategy selection, and confidence thresholds based on what the market is doing right now.

market_regime: BULL_CALM

SPY  +0.4%  ███████

QQQ  +0.7%  █████████

DIA  +0.2%  ████

VIX  14.2   low volatility

→ full position sizing enabled

→ momentum strategies active

Risk is architecture, not a setting

Five layers of protection run on every single trade. Not optional. Not configurable to "off." The engine enforces risk limits at the code level, not the UI level.

  1. Bracket orders (SL + TP)

  2. Max position count

  3. R-based position sizing

  4. Daily loss limit auto-kill

  5. EOD liquidation

all layers: ENFORCED

ui_toggle: NONE

An autonomous trading agent that explains every decision, enforces its own limits, and gets smarter with every trade. Not a black box. Not a dashboard of settings. A reasoning system you can actually audit.

@saebl_app

THE DECISION LOOP

Fully autonomous.

Observe → Reason → Decide → Execute → Learn → Repeat

01

Observe

Scans prices, technicals, regime signals, and news across your watchlist every cycle.

02

Reason

Analyzes the data, considers past lessons, and documents its reasoning for audit.

03

Execute

Validated decisions become bracket orders with stop-loss and take-profit built in.

04

Learn

Every closed trade is analyzed. Lessons extracted, evidence tracked, playbook updated.

05

Repeat

Next cycle starts with everything learned. Better data, better rules, better decisions.

THE DASHBOARD

Everything. One screen.

Real-time P&L, open positions, market regime, engine status, learned strategies, and trade history — all in one place. No tab-switching. No second app.

Saebl
Engine Running
SESSION +$2,847

REGIME

BULL_CALM

WIN RATE

67%

DRAWDOWN

-2.1%

POSITIONS

3 / 5

MARKET

02:14:37

SPY +0.8% QQQ +1.2% DIA +0.3% VIX -4.2%

OPEN POSITIONS

NVDA Long 15 @ $177.40
+$390
AAPL Long 25 @ $260.50
+$670
MSFT Long 10 @ $406.30
-$230

AGENT LOG

▲ BUY
NVDA 15 @ $179.40
85%

└ breakout + volume spike

◼ HOLD
AAPL — consolidating
BULL_CALM (0.82)

Saebl is in pre-launch. We'll be publishing real trading data soon.

WHO IT'S FOR

Traders who've been burned by black-box bots and want to understand what's actually happening.

You know the markets but can't watch screens all day. You want an agent that runs autonomously while you do other things — but you need to be able to audit every decision it makes.

You care about downside protection first. Five layers of mandatory risk controls mean the engine kills itself before one bad day wipes out your month.

You don't trust tools you can't inspect. Saebl shows every decision trace, every lesson, every reasoning chain. Read the logs. Verify the logic. Audit the pipeline.

Get early access

Saebl is launching soon. Be first in line.

PRODUCTION GRADE

Not a prototype. Not a demo.

Atomic writes

Session state uses tmp+rename for crash-safe persistence. No half-written files. No corrupted state. Ever.

Pipeline tracing

Every decision, execution, and learning event is logged to an append-only trace file. Full audit trail.

Crash recovery

If the engine crashes, it recovers open positions, catches up on learning, and shows you what happened.

Async everything

All broker and LLM calls are async with timeouts and retries. The event loop never blocks.

LLM validation

Every LLM response is validated before acting. Malformed outputs get safe defaults, not crashes.

Desktop native

Electron app with auto-update, backend auto-restart, and local-first data. Your machine, your keys.

Stop watching.
Start deploying.

An autonomous trading agent that reasons, executes, and learns.
Local-first. Production-hardened. Your money, your machine, your rules.

Read the Architecture