Home / LLM agents & research
Xtra-Computing/CryptoTrade
Reflective LLM agent for zero-shot crypto trading
What it is
Research code for CryptoTrade, a reflective LLM-based agent that guides zero-shot cryptocurrency trading, published at the EMNLP conference. The agent combines on-chain and market data with a reflection step that reviews its past decisions, without fine-tuning for specific coins. Researchers and developers studying LLM agents for trading are the intended audience. It is paper code, so it supports reproducing the study rather than running a live trading service.
At a glance
Research onlyOur rating, based on popularity, maintenance and how ready it is for real use.
| Best for | Professional quants |
|---|---|
| Used for | Strategy research, AI agents |
| Markets | Crypto |
| Stack | Python |
| Learning curve | Moderate learning curve |
| Practical value | Medium practical value |
| Cost | Free and open source |
| Hardware | No GPU needed |
| Maintenance | No commits in over six months |
GitHub stars, last 30 days
Daily snapshots since 2026-09-28 (up to 30 days): 0 over the period, now 166. Gaps mean no snapshot was taken that day.
[EMNLP 2024] CryptoTrade: A Reflective LLM-based Agent to Guide Zero-shot Cryptocurrency Trading https://aclanthology.org/2024.emnlp-main.63.pdf
Similar tools
LLM trading research agent with multi-market backtesting
Multi-agent LLM framework for crypto trading decisions
Open-source LLM harness for agentic trading
Deep reinforcement learning agent that trades Bitcoin
Multi-agent LLM research system for China A-shares
Tutorial project building a reinforcement learning Bitcoin bot