Repo

circuit-framework

Generates crypto trade proposals from multi-agent analysis, then sizes, clamps, or rejects them with deterministic risk rules before paper execution.

GitHub ↗Jul 16, 2026Paper ↗

  • llm
  • trading-agents
  • paper-trading
  • risk-engine
Pipeline diagram showing multi-agent trading system from data to execution
Pipeline diagram showing multi-agent trading system from data to execution

What it is

Multi-agent crypto research and paper-trading framework that routes LLM trade ideas through a deterministic risk gate before any simulated execution. Run crypto market analysis with specialized LLM agents, produce structured LONG/SHORT/NO_TRADE proposals, and paper-trade only after deterministic risk checks approve them.

Key results

  • Uses only Hyperliquid public Info API; no auth or wallet credentials
  • Supports LONG, SHORT, and NO_TRADE; NO_TRADE is never converted to a fill
  • Offline crypto unit tests use fixtures/mocks and must not hit the network

Shows LLM trading research can be constrained by a deterministic approval layer and kept fully paper-only.

One useful ML email a week.

Worth your time, not all of your time.

We track new ML work across papers, code, and data, then keep only the pieces worth a closer look. Selected, not collected.

Papers.Data.Code

The Weekly ML Signal

Selected, not collected.

One useful ML email a week: the papers, repos and datasets worth your attention.

Worth your time, not all of your time.