The Trading Agentic Stack, Decomposed and Curated

What it does awesome-trading-agents is a curated index of open-source projects where LLMs research markets, make trading decisions, or wire agents to market data. It deliberately excludes classic quant libraries and reinforcement-learning bots, pointing readers to older awesome-lists for those. Entries are vetted for public code, clear LLM-driven behavior, recent activity, and visible adoption.

The interesting bit The list imposes a three-layer taxonomy—Agents, MCPs, and Skills—that mirrors how these systems are actually built. Agents are the decision-makers, MCPs are the data and execution plumbing, and Skills are reusable prompts or capabilities. This structure turns a noisy gold rush into a parts catalog, making it easier to find, say, a multi-agent debate framework or a Kraken MCP server without wading through unrelated backtesting tools.

Key highlights

• 545 stars, maintained by the LLMQuant community with bilingual English and Chinese coverage.
• Three clear sections: Agents (e.g., TradingAgents, ai-hedge-fund), MCPs (e.g., alpaca-mcp-server), and Skills.
• Explicitly excludes traditional quant/RL tools; scope is strictly LLM-native behavior.
• Selection criteria include public artifacts, useful documentation, distinct role, and recent activity checks.
• Released under CC0-1.0, so the index itself is free to reuse.

Caveats

• It is a directory, not a framework—expect links and descriptions, not a unified SDK or installable stack.
• Dates are not shown after every item, so you will need to verify freshness yourself.
• No explicit maturity tiers or quality rankings beyond the maintainers’ selection criteria.

Verdict Bookmark this if you are building or evaluating LLM-native trading infrastructure and need to survey the landscape. Skip it if you want a single turnkey trading platform or classical quantitative research tools.