cybolabs
AI trading platform Multi-tenant Containerized

One platform: research, simulate, execute, observe.

The cybolabs AI trading platform is a composable infrastructure for trading businesses: a multi-tenant broker-dealer core, a clustered historical data store, a live market-data fan-out, and client + operator UIs - with AI assistance layered on top.

Request access White-label it Try the demo

The loop

Every stage is observable and replayable; nothing goes live without passing the previous gate.

[ research ] -> [ backtest ] -> [ paper ] -> [ live ] -> [ telemetry ] ^ | +------------------- feedback -----------------+

What's in the platform

Broker-dealer core

Multi-tenant execution plane: per-client accounting, risk, compliance, credit/margin, in-house clearing, and order routing. Serves brokers, individuals, and agentic clients.

  • Per-tenant ledger, exact to the cent
  • Pre-trade risk checks and position limits, non-bypassable
  • Credit/margin and in-house clearing
  • Order routing across venues

Historical data server

A clustered OHLCV bar store with replication. A single source of truth per bar; the primary owns writes, replicas subscribe.

  • Clustered, replicated bar storage
  • Single source of truth per bar
  • Replay-grade historical access

Market data service

Live market-data fan-out from multiple sources. Normalized quotes, clock, and snapshots over REST and WebSocket - push-only.

  • Multi-source, normalized quotes
  • REST + WebSocket, push-only (no polling)
  • Snapshots and a coherent clock

Client + operator UIs

Two separate sites with a hard wall between them: a public client site and a VPN/IP-ACL operator site. Identity via per-realm SSO with MFA on sensitive actions.

  • Dashboard, orders, chart, positions, blotter, statements
  • Strategies, replay/backtest, risk, API keys, webhooks
  • Hard client/admin isolation
  • MFA for live switch and API key issuance

Operating principles

P1

Determinism first

Same inputs, same outputs. Backtests and replays are byte-comparable.

P2

Risk before speed

Pre-trade checks and limits are non-bypassable; speed never overrides safety. Breaches surface in a live feed.

P3

Observe everything

Live data is push-only over WS; cross-service events fan out over a shared message bus. Every order, fill, and decision is logged with enough context to replay the day.

Engineering

Stack

Python / FastAPI / Pydantic v2 / async data layer / Redis / Postgres / Docker Compose. Contract-first OpenAPI 3.1.

Composable

Each service is independently deployable and replaceable as long as it honours its published contract. One shared network, one decisions log.

Self-hostable

Containerized; your infrastructure, your keys, your data. External venue/market-data connectivity is gated and operator-approved per environment.

Talk to us about access