Every entry is a plain file in a public git repository — validated against open schemas, connected in a typed knowledge graph, and readable by humans and agents alike. Share the evidence, not the edge.
Browse the registry
- AI trading agents and agent systems.
- Design patterns and frameworks for building agentic traders.
- Data for training and evaluating trading agents.
- Reusable capabilities for coding, research, and trading agents.
- Research, connected to implementations and reproductions.
- Standardized evaluations for trading agents and financial models.
- Results ordered by verification and track length — never raw return.
Filter everything by type, market, level, and tag →
Recently added
The reference passive baseline every active result should be compared against; buys the index and does nothing.
passive-baselinefully-openequitiesadded 2026-08-15 · MIT · native- FinMemAgent
An LLM trading agent that combines a configurable character profile with a layered memory module modeled on human trader cognition.
fully-openequitiesai: advancedsetup: advancedadded 2026-08-15 · MIT · external - FinRobotAgent
An open-source multi-agent platform from AI4Finance Foundation that applies LLM agents to equity research, trading strategy, and risk evaluation tasks.
autogenlangchainfully-openequitiesadded 2026-08-15 · Apache-2.0 · external - FreqtradeAgent
A Python cryptocurrency trading bot with backtesting, hyperparameter optimization, and an adaptive machine-learning module called FreqAI.
fully-opencryptoadded 2026-08-15 · GPL-3.0 · external - HummingbotAgent
An open-source Python framework for building and running market-making and arbitrage bots across centralized and decentralized crypto venues.
market-makingarbitragefully-opencryptotrading: advancedadded 2026-08-15 · Apache-2.0 · external - TradingAgentsAgent
A multi-agent LLM framework that assigns specialized analyst, researcher, trader, and risk-management roles to simulate the workflow of a trading firm.
langgraphfully-openequitiesadded 2026-08-15 · Apache-2.0 · external
Share the evidence, not the edge
Publishing here never requires giving away a strategy. Results are welcome at every disclosure level, and every one carries its mechanism on its face:author-reported is the honest default, not a failure. Negative results are first-class citizens — a documented loss teaches more than an unverifiable win.
Report a result








