gerskill/moondev-trading-ai-agents

Purpose and Entry Points

This is an experimental Python research system for AI-assisted crypto trading analysis, monitoring, risk checks, strategy research, and optional trade-decision workflows. The primary runtime entry point is src/main.py, which enables or disables agents. Other operational surfaces include a bot-oriented runner at src/ezbot.py, an API module at src/agents/api.py, and a separate frontend application at src/frontend/main.py.

Agent Runtime and Shared Contract

Agents are organized as independently runnable domain workers under src/agents/, with common behavior likely centralized in src/agents/base_agent.py. This boundary allows market-monitoring, analysis, risk, and research capabilities to evolve separately while being orchestrated from the main runtime.

• src/agents/base_agent.py

• src/agents/__init__.py

• src/main.py

Trading, Strategy, and Risk Decision Chain

The core decision-oriented path combines token/trade analysis, strategy execution or approval, and portfolio safety controls. The trading agent produces basic trade decisions from token data; the strategy agent manages strategies placed in the strategies package; the risk agent enforces position and PnL/minimum-balance constraints. Risk is an important boundary before any execution-oriented behavior.

• src/agents/trading_agent.py

• src/agents/strategy_agent.py

• src/agents/risk_agent.py

Strategies implement a separate extension layer, anchored by a base type and examples/custom strategy location rather than being embedded in agents.

• src/strategies/base_strategy.py

• src/strategies/custom/example_strategy.py

• src/strategies/example_strategy.py

Market Intelligence and Data Collection

Monitoring agents ingest and interpret crypto market signals including funding, liquidations, open interest, whale activity, listings, and CoinGecko token information. They persist or consume local CSV histories, creating a file-backed state boundary rather than a dedicated database.

• src/agents/funding_agent.py

• src/agents/liquidation_agent.py

• src/agents/whale_agent.py

• src/data/funding_history.csv

• src/data/liquidation_history.csv

• src/data/oi_history.csv

The README identifies a Moon Dev API v2-style market-data integration for funding, liquidation, open-interest, and CopyBot data, plus CoinGecko and Hyperliquid/Solana-oriented workflows. Required secrets include Anthropic and other trading/API keys.

AI Analysis and Notification Agents

Several agents apply LLM analysis to distinct inputs: charts, Twitter sentiment, CopyBot activity, and token listings. Outputs include buy/sell/nothing recommendations, saved analyses, historical sentiment, and voice alerts. LLM-provider configuration is centralized at the application level.

• src/agents/chartanalysis_agent.py

• src/agents/sentiment_agent.py

• src/agents/copybot_agent.py

• src/agents/listingarb_agent.py

• src/config.py

Tweet generation and sentiment-related artifacts are persisted locally.

• src/agents/tweet_agent.py

• src/data/sentiment_history.csv

• src/data/tweets/

Arbitrage and Research Pipelines

Specialized agents target funding-rate arbitrage and RBI strategy research. The RBI flow takes a strategy description/video/PDF concept, uses DeepSeek for research, and produces or supports coded backtests. Research inputs, generated strategy descriptions, backtests, and packaged backtests live together under data/rbi.

• src/agents/fundingarb_agent.py

• src/agents/rbi_agent.py

• src/data/rbi/research/

• src/data/rbi/backtests/

• src/data/rbi/backtests_package/

User/API Boundary and Infrastructure

The frontend is a standalone Python-served UI with HTML, JavaScript, CSS, and static assets. Programmatic access is exposed separately through the agents API module. The repository is Python-based with dependencies declared in requirements.txt; operational data is primarily CSV and local files.

• src/frontend/main.py

• src/frontend/templates/index.html

• src/frontend/static/js/main.js

• src/agents/api.py

• requirements.txt