Discover free. Prove before you deploy.

Every plan connects your AI agent and includes the full Feature Lab. Upgrade when you need more data, stronger evidence against overfitting, and reinforcement learning.

No credit card for Free · Cancel paid plans anytime · Your data stays local

Free

Discover a signal

Get the first evidence

Starter

Research deeper

Expand data and history

Pro

Prove & train

Attack overfit and learn

01 · Discover

Free Sandbox

Discover promising signals on your own data before paying

$0/ month
  • Explore features on up to 10K bars
  • Local AI agent access over MCP
  • Event-driven OHLC backtesting
  • Feature Lab: linear + neural + probes
  • Core performance evidence
  • Community Discord support

02 · Research deeper

Starter

Expand promising research across 10× more market data

$19/ month
  • Explore features on up to 100K bars
  • Local AI agent access over MCP
  • Event-driven OHLC backtesting
  • Feature Lab: linear + neural + probes
  • Saved reports & dashboard metrics
  • Core performance evidence
Overfit proof + RL

03 · Prove & train

Pro Agent

Prove robustness, optimize at scale, and train trading models

$49/ month
  • Local AI agent access over MCP
  • Feature Lab: linear + neural + probes
  • Unlimited bars & backtests
  • Reinforcement Learning Engine (DQN)
  • WFA, Monte Carlo & sensitivity proof
  • Portfolio, intrabar & exit optimization
  • Parallel Grid Optimization (Rayon)
  • CSV, Excel, & Database report export
  • Up to 3 active device activations

Common to every plan: local MCP agent connection · full three-stage Feature Lab · event-driven backtesting · private on-device research

New · Feature Lab v3

Find useful inputs before you optimize a strategy around noise.

Bring RSI, order-flow data, calendar fields, or indicators with any name. RLXBT ranks what carries signal, checks whether a feature adds nonlinear value out of sample, and then demands evidence from a strategy that can actually trade it.

Neural evidence is treated as a lead, not proof. Promotion requires stable folds and executable strategy evidence.

01

Screen the signal

Quality, information coefficient, stability, and response profiles expose weak or leaky inputs early.

02

Test nonlinear value

A challenger neural model must beat the price-only baseline across expanding out-of-sample folds and seeds.

03

Demand tradable evidence

Generated strategy probes check whether the discovered pattern survives rules, costs, and real execution logic.

Compare by outcome

What changes when you upgrade?

The agent and Feature Lab stay. Your research depth and standard of proof increase.

Research capabilityDiscoverFreeResearchStarterProve & trainPro
MCP agent + full Feature LabIncludedIncludedIncluded
Bars per backtest10K100KUnlimited
Saved visual research workspaceIncludedIncluded
Walk-Forward + Monte CarloIncluded
Sensitivity + portfolio + intrabarIncluded
DQN reinforcement learningIncluded
Active devices113

Frequently Asked Questions