Distributed intelligence for the futures market
Plexus is an open protocol for AI trading strategies. Like an octopus, the intelligence isn't locked in one head — it flows through a living network to every arm: the market is read, refracted into a decision, and fired out to your trading platform.
The framework and the plumbing are open source and free — transparency is the product. The proven, premium AI models are available at PlexusTraders.com.
flowchart LR
M["Market data<br/>raw + noisy"] --> P["PRISM<br/>the brain · refract → decision"]
P -->|Plexus Bus| NT["NinjaTrader"]
P --> MT["MetaTrader"]
P --> QT["QuantTower"]
P --> API["Direct API"]
P --> AX["Axon<br/>our headless ML client"]PRISM is the brain at the center — every trade client branches from it over the Plexus Bus. Axon is our own premium, headless client (more below).
Why we open the framework
We open the plumbing — the protocol, the bus, the connectors, the plugin system — and keep the alpha sealed. We're not handing out our winning models; we're giving the whole trading community better rails to build on.
- Contributing to trading. Real, production-grade infrastructure — an open protocol and tools any developer or trader can build on, not a toy.
- Helping you win. Connect your platform, write your own AI plugins, and run them on the exact rails we trade on. The framework makes you more capable, for free.
- Trust through transparency. The space is full of black boxes with hidden pricing and unverifiable "80% win-rate" claims. Read the protocol and run the tools yourself — so by the time you reach the premium models, you already know the system is real.
What stays private is the edge: the sealed Axon models at PlexusTraders.com. Open system, private alpha.
-
An open protocol
The Plexus Bus wire protocol, three conformant language implementations, and a public spec — the nervous system that connects every component.
-
Bring your own strategy
Write a Python plugin against the PrismR daemon and tap the live bus — forecasting, ML, your own logic, running inside the free monitor.
-
Feed any platform
Start with NinjaTrader; the engine is platform-agnostic by design. Every new platform is the same intelligence reaching a new market.
-
Premium Axon models
When you want proven, automated edge: sealed, ML-trained Axon models, transparent pricing, no black boxes.
Built for high-frequency trading
Plexus speaks a custom binary wire protocol. On the hot path it negotiates MessagePack with a content-addressed schema registry — a market-data bar drops from ~250 bytes of JSON to ~40 bytes, around 80% smaller. Across a 5-instrument feed at 10 bars/sec that's 540 KB → 86 KB per minute: the bandwidth and latency headroom that order flow and high-frequency execution demand. Every message is HMAC-signed, with an optional sealed transport for a fully closed bus.
We proved the system in-house on a Python stack — the PRISM coordinator plus our Ammonita engine, which is what we trade on today. The Rust rewrite (PRISM + the sealed Axon engine) is the next level: roughly 10× the throughput, built for low-latency high-speed trading.
The pieces, named after the system
| Plexus | the nervous system | the open bus/protocol connecting every component |
| PRISM | the brain | refracts raw, noisy market data into a clear decision |
| Axon | the execution arm | our proprietary headless trade client (like NautilusTrader) with a built-in ML engine — the premium one |
The brain decides; the trading platform is just the hand that executes.