Engineering Behind
the Speed
C++ execution core, Go orchestration layer, ZeroMQ message bus — every component built for deterministic, sub-millisecond performance.
System Architecture
Three layers, one message bus. Every component communicates over ZeroMQ — no HTTP overhead, no polling, no wasted cycles.
MT5
Gateway
Execution Engine
Bridge Service
FIX
4 LPs Connected
FIX Protocol Engine
Written in C++ for deterministic execution. Handles session management, sequence recovery, heartbeats, and reconnection — all at wire speed.
FIX 4.2 & 4.4
Full support for both protocol versions. Tag-level customization per LP — handle non-standard tags, custom exec reports, and vendor-specific extensions.
Session Management
Automatic logon/logout, heartbeat monitoring, test requests, and gap fill. Every session self-heals on disconnect with sequence-aware reconnection.
Zero-Copy Parsing
FIX messages parsed in-place without memory allocation. Tag extraction, validation, and routing happen on the original buffer — no copies, no GC pauses.
Multi-Session
Hundreds of concurrent FIX sessions to different LPs. Each session runs on its own thread with dedicated socket — no contention, no head-of-line blocking.
Sequence Recovery
Persistent sequence numbers with gap detection. On reconnect, automatically requests resend of missed messages — no lost orders, no duplicate fills.
Message Logging
Every FIX message logged with microsecond timestamps. Full audit trail for compliance, debugging, and post-trade analysis — searchable through the dashboard.
Execution Flow Engine
Visual block-based architecture where every order follows a dynamic path. No fixed pipelines — only the blocks that apply get executed.
Block Architecture
Each execution step is a self-contained block: A-Book, B-Book, Split, Delay, AI Score, Last Look, Markup, Volume Filter. Blocks wire together in any order — the engine evaluates the graph at runtime.
Conditional Routing
Orders route through different paths based on real-time conditions: account risk score, symbol, volume, time of day, LP availability, spread threshold. Every condition is a configurable rule, not hard-coded logic.
Short-Circuit Evaluation
If a block determines the order doesn't match its conditions, the entire downstream branch is skipped. No wasted evaluations — orders take the shortest possible path through the flow.
Hot Reload
Flow changes apply instantly without restarting the engine. Edit a routing rule, add a new block, rewire a path — all live. Orders in-flight continue on their original flow; new orders pick up the change.
Price Aggregation
Real-time best bid/ask construction from multiple LPs with markup, spread enforcement, and depth of market — all computed in the C++ core.
Multi-LP Aggregation
Prices from all connected LPs merged into a single composite book. Best bid/ask selected per symbol with configurable depth levels.
Markup Engine
Per-symbol, per-group markup rules. Fixed spread, percentage, or tiered markup based on volume. Applied in real-time before price distribution.
Spread Enforcement
Minimum spread floor per symbol. If LP spreads tighten below your threshold, the engine automatically widens to protect margin.
Stale Price Detection
LP prices that haven't updated within a configurable window are automatically excluded. No stale quotes reach your traders.
Price Filtering
Spike detection, outlier rejection, and band filtering. Anomalous ticks from LPs are caught and discarded before aggregation.
ZMQ Distribution
Aggregated prices published over ZeroMQ PUB/SUB. Subscribers receive only the symbols they need — no bandwidth waste, microsecond delivery.
AI Intelligence Engine
Machine learning models that profile every trading account in real-time. Score, classify, and route — automatically.
Trader Profiling
Every account scored 0–100 based on trading patterns: win rate, hold time, trade frequency, slippage capture, and order-to-fill ratio. Models retrain hourly on live data.
Flow Classification
Automatic detection of toxic flow, scalpers, HFT, news traders, and arbitrageurs. Each classification maps to a routing strategy — A-Book toxic, B-Book retail, split the rest.
Score-Based Routing
Execution flows branch on AI scores. High-risk accounts route to different LPs, get wider spreads, or trigger delay blocks. Low-risk accounts get best execution. All configurable per threshold.
Self-Learning Models
Models adapt as trader behavior changes. An account that was retail last month but turned toxic this week gets reclassified automatically — no manual intervention required.
MT5 Integration
Native gateway integration with MetaTrader 5. Orders, positions, and account data flow seamlessly between your trading servers and the bridge.
Gateway API
Direct integration via MT5 Gateway API. No plugins, no DLLs — native server-to-server communication with full order lifecycle support.
Multi-Server
Connect multiple MT5 servers to a single ZeroMS instance. Each server operates independently with its own routing rules and LP assignments.
Position Sync
Real-time position synchronization between MT5 and LP hedging accounts. Open, modify, close — every action reflected instantly on both sides.
Group Mapping
MT5 groups map to bridge taker groups. Different execution flows, markups, and LP pools per group — configure once, apply to thousands of accounts.
Symbol Mapping
Flexible symbol translation between MT5 and LP conventions. EURUSD → EUR/USD, XAUUSD → GOLD — all handled transparently by the bridge.
Event Streaming
Every MT5 event — deal, order, position change — streamed over ZeroMQ in real-time. The dashboard, alerts, and risk engine all subscribe to the same feed.
The Stack
Three technologies. No frameworks, no abstractions, no overhead.
Execution Core
- FIX protocol engine
- Price aggregation
- Order matching & routing
- Session management
- Zero-copy message parsing
Bridge API & Orchestration
- REST API layer
- Configuration management
- Flow engine & rule evaluation
- Risk calculation
- Alerting & notifications
Message Bus
- PUB/SUB price distribution
- REQ/REP order execution
- PUSH/PULL event streaming
- Inter-process communication
- Microsecond message delivery
See the engine
in action
Schedule a technical deep-dive. We'll walk through the architecture, execution flows, and AI engine with your team.