Multi-bot maker-only grid swarm engine with fee-aware adaptive sizing, evidence-locked accounting, circuit breakers, drawdown modes, and a live observability dashboard. Built for systematic researchers who need full visibility and full control. Research and simulation tool only—not financial advice.
Volatility-harvesting grid strategies on crypto markets require careful fee accounting, regime awareness, and precise grid discipline to operate without destroying edge through taker fees or excessive churn.
This system provides the research and simulation infrastructure to study, calibrate, and observe grid swarm behavior—with a full governance stack: circuit breakers, kill switches, drawdown modes, and an evidence-locked SQLite accounting layer that makes every decision reproducible.
Strategy
Maker-only grid
Architecture
Multi-bot swarm
Fees
Fee-aware adaptive
Accounting
SQLite truth layer
Safety
Circuit breakers
Observability
Live dashboard
Multiple independent grid bots operate across configurable pairs and parameters. Each bot operates within its own allocation, contributing to aggregate behavior that the operator monitors and controls from a single observability layer.
All orders are placed as limit orders at the maker side of the spread. The system enforces maker-only behavior by design—taker fills are logged as anomalies, not accepted as normal. This discipline is the core fee-protection mechanism.
Grid spacing and order sizing are adapted based on current fee structure and spread conditions. Grids that would generate negative expected value after fees are flagged or deferred—not executed.
Every order placement, fill, cancellation, and fee event is written to an append-only SQLite truth layer. Reproducible "effective config" snapshots capture the exact parameter state at any point, making any past behavior fully reproducible.
Drawdown triggers, churn rate anomalies, fill efficiency degradation, and regime shift detectors can pause individual bots or halt the entire swarm. A single operator kill switch stops all activity immediately.
Real-time metrics: live P&L by bot, grid fill rates, reject codes, churn/fill efficiency, open order depth, and exposure vs. allocation targets. Operators see everything; the system hides nothing.
Each bot operates within hard allocation limits. Aggregate exposure caps prevent the swarm from collectively exceeding configured risk parameters even as individual bots operate independently.
Volatility regime detection determines when grid strategies are favorable vs. adverse. Regime gates defer or halt bot activity during conditions the strategy is not designed for—rather than operating blindly.
Configurable drawdown thresholds trigger defensive modes: reduced grid density, widened spacing, or full pause. Drawdown mode transitions are logged with timestamp and the metric that triggered them.
At every parameter change, the system captures an "effective config" snapshot. Any historical period can be replicated by restoring the snapshot active at that time—making retrospective analysis exact.
Order rejections are categorized by reason code. Churn rate (cancellations per unit time) and fill efficiency (fills per placement) are surfaced as health metrics. Degradation triggers alerts before it becomes material.
No parameter changes, bot additions, or strategy modifications occur without explicit operator action. The system does not self-modify, self-optimize, or operate outside its configured constraints.
The kill switch, circuit breakers, and evidence-locked accounting layer were designed before the grid logic. Control architecture is not optional — it is the foundation. Read our Governed intelligence, not guesswork governance framework →
Researchers studying maker-rebate strategies, grid behavior across volatility regimes, and fee impact on grid profitability who need a rigorous, evidence-locked research environment.
Developers building and testing grid execution logic who need reproducible simulation environments, exact config snapshots, and observable system state at every decision point.
Operators who understand that a well-constrained system with clear circuit breakers is safer than an unconstrained system with confident-sounding output. Control is the feature.
We'll walk through the architecture, the accounting layer, and the governance stack—and discuss what a deployment scoped to your research context looks like.