Autonomous Trading Ecosystem (CONCEPT:AU-ECO.bus.pluggable-event-queue)

Overview

The Autonomous Trading Ecosystem provides a full suite of infrastructure, configurations, and SI-native patterns for executing quantitative and algorithmic trading via the emerald-exchange MCP server and the agent-utilities ontology framework.

1. Unified Configuration Schema

The trading ecosystem is configured centrally in the XDG-compliant ~/.config/agent-utilities/config.json file under the "trading" root key. It supports: - Exchange Backends: Abstractions for paper, alpaca, ccxt (Binance, Coinbase, Kraken), and freqtrade. - Risk Limits: Automated OS-5.1 pre-trade controls (max_position_pct, max_portfolio_drawdown_pct, regime_shift_halt). - Signal Fusion: Configurations for bayesian fusion and minimum confidence scoring. - Monitoring Intervals: Background cron scheduling frequencies for P&L tracking and snapshotting.

2. Infrastructure Blueprints

Automated provisioning of trading nodes is handled via the skill-graphs/infrastructure-blueprints/trading/ YAMLs: - emerald-exchange.yaml: The primary MCP server handling order execution and pre-trade risk validation. - freqtrade-node.yaml: A pre-configured Freqtrade container for crypto-native algorithmic execution. - trading-grafana.yaml: Observability stack deploying Grafana + Redis TimeSeries for portfolio visualization. - gpu-compute-node.yaml: Dedicated GPU node routing to data-science-mcp for QLib backtesting and model training.

3. Skill-Graph Decomposition (CONCEPT:AU-AHE.assimilation.trading-ecosystem-spec)

To enable multi-agent specialization, trading knowledge is broken down into four discrete skill-graphs subdirectories: - strategies/: Documentation of momentum, mean reversion, multi-factor, and ML trading archetypes. - data-sources/: Market data ingestion patterns (AKShare, Yahoo Finance, CCXT) and on-chain analytics. - execution/: Backend protocol implementation details and the order lifecycle. - risk-management/: Implementation of Kelly criterion sizing, KS-Test regime detection, and Kill Switches.

4. MCP Domains

The emerald-exchange MCP server exposes 9 discrete tool domains: 1. market_data: Historical OHLCV and real-time quotes. 2. orders: Submit, cancel, status, halt, resume. 3. portfolio: Account equity, positions. 4. risk: Pre-trade validation, daily loss checks, Kelly sizing. 5. signals: Alpha generation, signal fusion, regime detection. 6. strategy: Lifecycle management (draft → paper → live). 7. crypto: Funding rates, on-chain whale alerts, cross-exchange arb scanning. 8. debate: Multi-agent hypothesis debates with risk compliance veto power. 9. prediction_markets: Prediction-market tooling with risk-guard integration.

Implementation Details

• Source Code: agent-packages/agents/emerald-exchange
• Pillar: Ecosystem Peripherals
• Bridge: CONCEPT:AU-AHE.assimilation.autonomous-trading-ecosystem through EE-020