VVhalanx
Autonomous multi-agent trading · governed by RL

A trading desk that runs itself — with a risk officer that never sleeps.

Vhalanx runs a book of specialist trading agents across day-trade and swing horizons. Every order they propose is judged in real time by a reinforcement-learning governor before it ever reaches a broker — and the same engine is multi-tenant, so anyone can run their own agents on it.

Dual-broker paper pilot · since 2026-05-13 First real-money sleeve · LIVE 5 beta tenants onboarded
vhalanx · execution pipeline SESSION LIVE
01 · ingest Market Data IBKR + SIP feeds over a Kafka plane. One publisher, every tenant.
02 · strategy Agents Momentum, trend, mean-revert, gap, breakout, sniper — each proposes trades.
03 · risk gate RL Governor ● ALLOW 1.0×
04 · execute Broker Sized off broker truth. Trailing stops, caps, and kill-switch applied.

↑ A live simulation of the decision loop. Agent proposals stream in; the governor scales, throttles, or halts each one.

The core idea

Strategies find the trade. A learned governor decides if it survives.

Most trading bots are a single strategy with a fixed rule book. Vhalanx separates finding edge from managing risk — and puts a trained model in charge of the second job.

Separation of duties

Many narrow agents

Each agent does one thing well and stays honest about its edge. No single strategy carries the whole book, so no single failure sinks it.

The differentiator

An RL risk officer

A PPO-trained governor sits between every agent and the broker. It reads regime and exposure and scales each order — allow, throttle, or halt — the way a human risk desk would.

Truth, not counters

Broker-anchored

Sizing, P&L, and stops read live broker state, not internal tallies. What the account actually holds is the source of truth end to end.

Why it matters: the governor is trained and validated separately from the strategies, and generalizes out-of-sample. It's the piece that turns a pile of signals into a book you'd trust with real money.
Inside the engine

Two horizons, one governed pipeline.

The same architecture runs fast intraday strategies and multi-day swing positions — the horizon changes, the governance doesn't.

Day-trade book

Intraday specialists

  • Sniper — precision breakout entries, bracketed on entry
  • Gap — opening-range continuation
  • Momentum & breakout — trend-continuation on liquid names
  • Rider — manages open positions, trailing exits, end-of-day risk-off
Swing book

Multi-day positioning

  • Momentum & trend — regime-aware holds; stand down when the broad trend rolls over
  • Mean-revert — counter-trend reversals on stretched names
  • Drawdown-locked exits — trailing stops that ratchet up with the trade
  • Staged profit-taking — a seatbelt, not the edge
Risk controls, always on

Guardrails wrap every order

Independent of any agent's opinion, the engine enforces its own limits — and an operator can override all of it in one click.

Concentration capsPer-symbol and cumulative exposure ceilings
Regime gateStand down on new risk when the broad trend rolls over
Kill-switch / flattenNuke, pause, or flatten the whole book instantly
Max-hold + trailing stopsTime and drawdown limits on every position
The operating view

Every strategy runs as an isolated sleeve.

Capital is split into independent sleeves — by broker and by horizon — each with its own risk envelope, its own P&L, and its own record. Nothing is ever commingled.

+11.0% IBKR swing sleeve · paper pilot · NetLiq, indexed to 100

Endpoint is the actual broker-truth result: +$19K / +11% NetLiq on the paper IBKR swing sleeve since 2026-05-13. The intermediate path is illustrative — a weekly index, not a tick-by-tick reconstruction.

Not just our book — a platform

The same engine runs anyone's agents, isolated by tenant.

Vhalanx is multi-tenant from the ground up. A new user brings their own account, picks their own universe, and gets the full engine — data feed, agents, governor, and controls — walled off from every other tenant.

Onboarding

Bring your own broker

A tenant connects an Alpaca or IBKR account and configures the symbols they want to trade. The engine syncs their universe on the next cycle — no redeploy.

Data economics

One feed, many tenants

Free-tier users share a standardized pool broadcast over Kafka from a single market-data publisher. Paid tenants plug in their own SIP feed. Cost scales with symbols, not users.

Isolation

Walled-off books

Per-user universes, per-user P&L, and per-user state. One tenant's fills, positions, and controls never leak into another's dashboard.

Operator control plane

One switch reaches every engine

From the dashboard, an operator can pause, flatten, or nuke any sleeve or tenant — the command travels end-to-end to the live engine and takes effect on the next cycle. Loopback-only, defense in depth.

Attribution

Every fill traces to an agent

Fills carry a lifecycle ID and parent agent, so per-agent P&L is exact — and the governor's reward signal is de-censored, feeding the next round of training.

What's real today

Validated in backtest, proven on paper, live with real money.

The claims below are broker-truth or backtest results — labeled exactly as they stand. Paper is paper; only the ring-fenced sleeve is real money.

+11% NetLiq on the IBKR swing sleeve Paper pilot · broker truth · since 2026-05-13
$2K First real-money sleeve, live Ring-fenced · isolated stack · live 2026-07-08
20 yr Backtest battery on the swing edge +32% OOS · survivorship-clean · cost-sensitive
1.16 Governor Sharpe, out-of-sample Max drawdown ~22% · shadow-live on pilot
2 Brokers running the same agents Alpaca + IBKR · a live A/B, by design
5 Beta tenants onboarded Isolated books · shared data plane
The swing edge was stress-tested against a non-hindsight 2006 universe (survives), across regimes and walk-forward windows, and shown to be sensitive to costs — so the governor and execution layer exist precisely to protect the thin part of the edge.
Under the hood

Built to run unattended.

A Kafka data spine, a stateless engine that ships by immutable tag, and infrastructure defined as code — so the same system runs one book or many.

Data plane
Market-data publisher
IBKR gateway → Kafka
Confluent Kafka
topics, ACLs, archive
RL training corpus
every tick retained
Engine
vhalanx-core
agents + rider
RL governor
PPO risk gate
Broker adapters
Alpaca · IBKR
Control & view
vhalanx-api
control plane, loopback
Dashboard
per-sleeve, per-tenant
Attribution store
fill → agent lineage
Ship & run
CI → GHCR
deploy by immutable tag
Terraform / AWS
infra as code
Watchdog crons
auto-recover, self-heal
The pitch in one line

An autonomous trading engine you can trust — and a platform anyone can run it on.

Narrow agents find the edge. A learned governor keeps the book alive. Multi-tenant from day one, proven on paper, and already trading real money.