Overview
CTRLNET is an autonomous capital control network for creating, configuring, and supervising AI trading agents. It is a control plane—not a signal caller, not a guaranteed profit tool—but an infrastructure layer for agent registry, deployment orchestration, live monitoring, and policy-driven risk governance.
Users create strategy agents (Trend, Mean Reversion, Momentum, Arbitrage), configure risk profiles (Conservative, Balanced, Aggressive), allocate capital with granular controls, and monitor agent behavior through real-time telemetry. The system emphasizes governance over speculation: observe, constrain, supervise. This whitepaper defines the CTRLNET architecture, agent lifecycle, risk framework, and development roadmap.
Core Concepts
Agents
Autonomous software entities that execute trading strategies based on predefined logic. Each agent has a strategy type (e.g., Trend Following, Mean Reversion), risk profile (Conservative/Balanced/Aggressive), and allocated capital. Agents operate independently but are governed by system-wide constraints.
Strategies
Pre-built strategy templates define agent behavior. Trend strategies follow momentum signals. Mean Reversion strategies buy oversold conditions. Momentum strategies ride strength. Arbitrage strategies exploit price discrepancies. Users select templates and configure parameters.
Risk Profiles
Each agent operates under a risk profile: Conservative (low leverage, tight stops), Balanced (moderate exposure), or Aggressive (higher leverage, wider stops). Profiles enforce position sizing, stop-loss rules, and maximum drawdown limits.
Capital Allocation
Users allocate capital to agents with precision. Each agent has a max allocation. Global reserves are maintained. The system enforces capital limits and prevents over-allocation. Rebalancing is manual and audited.
Policy Engine
The policy engine enforces risk constraints: max allocation per agent, position limits, stop-loss requirements, circuit breakers. Policies are configurable but default to conservative parameters. Emergency overrides available.
Control Plane
The control plane is the oversight layer: agent registry, orchestration, monitoring, telemetry, audit logs. It provides visibility into agent status, decisions, and execution. The control plane does not execute trades—it supervises agents that do.
System Architecture
Agent Registry
Centralized registry of all agents: configuration, strategy, risk profile, allocated capital, deployment status. The registry serves as the source of truth for agent metadata.
Orchestration Layer
Manages agent lifecycle: creation, configuration, simulation, deployment, pause, resume, termination. Orchestration ensures agents adhere to policies during state transitions.
Risk Gate + Policy Constraints
The risk gate enforces global constraints: total capital exposure, per-agent limits, circuit breakers. If constraints are violated, agents are automatically paused. Emergency kill switch available.
Monitoring/Telemetry
Real-time telemetry streams agent heartbeats, status updates, decision events, and execution logs. Telemetry enables behavior analysis, anomaly detection, and performance review.
Audit Trail
Immutable log of all configuration changes, policy updates, capital allocations, and agent actions. Audit trail provides accountability and supports forensic analysis.
Agent Lifecycle
1. Create
User selects strategy template, defines agent name, and initializes configuration. Agent is registered but inactive.
2. Configure
User sets risk profile (Conservative/Balanced/Aggressive), allocates capital, and configures strategy parameters. Policy engine validates configuration.
3. Simulate
Agent runs in simulation mode using historical data. No real capital at risk. User reviews simulated behavior and performance metrics before deployment.
4. Deploy
Agent transitions to live mode. Capital is allocated, risk gate is engaged, telemetry begins streaming. Agent executes strategy according to configuration.
5. Monitor
User observes agent behavior via telemetry dashboard. Reviews decisions, tracks performance, checks policy compliance. Anomaly detection alerts on unusual activity.
6. Pause/Kill
User or risk gate pauses agent (temporary stop) or kills agent (permanent termination). Capital is returned to reserves. Audit log records the action.
Risk & Governance
Max Allocation Per Agent
Each agent has a capital ceiling. The system prevents over-allocation and enforces reserve requirements. Users cannot deploy agents without sufficient available capital.
Global Risk Gate
System-wide risk gate monitors total exposure across all agents. If aggregate risk exceeds thresholds, all agents are paused automatically. Risk gate settings are configurable but default to conservative limits.
Emergency Kill
Users can immediately terminate all agents via emergency kill switch. All positions are closed (if supported by strategy), capital is returned to reserves, and agents transition to killed state.
Conservative Defaults
CTRLNET defaults to conservative parameters: low leverage, tight position limits, mandatory stop-losses, frequent heartbeat checks. Users can override defaults but system encourages caution.
Roadmap
v1: Simulation Control Plane (Current)
- • Agent registry with strategy templates
- • Risk profile configuration and capital allocation
- • Simulation mode for testing agent behavior
- • Basic telemetry and monitoring dashboard
- • Policy engine with conservative defaults
v2: Real-Time Solana Telemetry Integration
- • Live Solana blockchain telemetry integration
- • Real-time position tracking and execution monitoring
- • Enhanced anomaly detection and alerting
- • Multi-agent coordination and conflict resolution
v3: Execution Adapters + Advanced Policies
- • Execution adapters for DEX integration (Jupiter, Orca, Raydium)
- • Advanced policy rules: time-based constraints, correlation limits
- • Cross-agent portfolio optimization
- • Governance tooling for multi-user deployments
- • Still emphasizing governance over guaranteed returns
Disclaimer
CTRLNET is a control plane for supervising AI trading agents. It is not financial advice. It is not a guaranteed profit tool. It is not a signal caller. CTRLNET is infrastructure for governance: creating agents, setting policies, monitoring behavior, enforcing constraints. Trading involves risk. Users are solely responsible for their capital, their strategies, and their outcomes. Simulation mode is recommended before deploying live agents. Always operate with capital you can afford to lose. No guarantees are made about performance, uptime, or profitability.