circuit-framework

Circuit Framework is a research tool that simulates cryptocurrency trading using a team of AI agents that debate with each other before deciding whether to make a trade. It is a fork of an earlier project called TradingAgents, adapted specifically for crypto markets. The README is very clear that this is for research only, not financial advice, and that it never places real trades or connects to a real exchange account.

When you ask it to analyze a coin like BTC or ETH, several specialized AI agents look at different angles of the market: price and order book structure, derivatives data like funding rates, social sentiment, news catalysts, and the overall market regime such as trend and volatility. Two more agents then argue the bullish and bearish case against each other, and a research manager weighs their arguments. A trader agent proposes a trade of going long, going short, or doing nothing, and a separate rule-based risk system checks that proposal against safety limits before deciding whether to approve it, size it, or reject it. Nothing is left purely to the AI's judgment at that final step.

Users interact with it through a command line tool, running commands like analyzing a symbol with a chosen strategy profile, or checking their simulated paper trading portfolio and position history. Several ready-made strategy profiles exist, such as balanced, momentum, and derivatives-focused, which change the agents' weighting without changing the underlying data pipeline. All trades in this framework are simulated on a local paper trading database, not real money.

The project supports many different AI providers behind the scenes, including OpenAI, Anthropic, Google, and several others, so users can choose which AI model powers the analysis. It also keeps the original stock market analysis mode from the project it was forked from, covering fundamentals, sentiment, news, and technical analysis for stocks. It is built in Python and licensed under the Apache License 2.0.