diff --git a/.lycheeignore b/.lycheeignore index 1b1942a..43e7c1d 100644 --- a/.lycheeignore +++ b/.lycheeignore @@ -2,3 +2,6 @@ # (client-side rendered + anti-bot). They work fine in browsers, so # the link target is valid for users — only the CI checker can't reach it. ^https://github\.com/topics/.*$ + +# Star History's image API intermittently returns 503 to CI clients +^https://api\.star-history\.com/svg\?repos=LLMQuant/awesome-trading-agents&type=Date$ diff --git a/README.md b/README.md index 10c5c16..f01382e 100644 --- a/README.md +++ b/README.md @@ -8,7 +8,7 @@

Awesome - GitHub stars + GitHub stars License: CC0-1.0 Topics Last commit @@ -83,6 +83,7 @@ Agents are projects where an LLM is part of the actual research or trading decis - [HKUDS/Vibe-Trading](https://github.com/HKUDS/Vibe-Trading) - Personal multi-agent finance workspace from HKUDS Lab; bundles Skills, MCP tools, and swarm presets across A-shares, HK, US, crypto, futures, and forex. - [brokermr810/QuantDinger](https://github.com/brokermr810/QuantDinger) - Open-source AI quant-trading platform; combines multi-agent research, backtesting, live trading, and multi-exchange routing. +- [Lumiwealth/lumibot](https://github.com/Lumiwealth/lumibot) - Backtestable AI trading-agent/team runtime with research, debate, risk, and memory in one backtest/paper/live strategy loop. - [The-Swarm-Corporation/AutoHedge](https://github.com/The-Swarm-Corporation/AutoHedge) - "Spin up an autonomous hedge fund in minutes"; applies the Swarms framework to market analysis, risk, and execution; CLI / SDK first. - [olaxbt/ai-market-maker](https://github.com/olaxbt/ai-market-maker) - Agentic crypto hedge-fund stack with specialist agents, Risk Guard, backtests, paper trading, and OpenClaw packaging. @@ -116,6 +117,8 @@ Agents are projects where an LLM is part of the actual research or trading decis - [virattt/ai-hedge-fund](https://github.com/virattt/ai-hedge-fund) - Widely forked LLM-driven equity-trading repo; analyst personas (Buffett / Munger / Cathie Wood) propose, the portfolio manager decides. - [TraderAlice/OpenAlice](https://github.com/TraderAlice/OpenAlice) - "Your one-person Wall Street"; single agent covering research → entry → hold → exit; Claude Agent SDK + Trading-as-Git approval workflow + cross-asset UTA account design. + +- [NoFxAiOS/nofx](https://github.com/NoFxAiOS/nofx) - Self-hosted LLM trading terminal; models decide and explain while a Go runtime enforces hard risk limits across nine exchanges. - [chrisworsey55/atlas-gic](https://github.com/chrisworsey55/atlas-gic) - General Intelligence Capital's self-improving trading agent; focuses on continuous self-research rather than agent debate. - [Gajesh2007/ai-trading-agent](https://github.com/Gajesh2007/ai-trading-agent) - AI trading agent on Hyperliquid; single-LLM-driven execution; shows direct Hyperliquid integration. @@ -145,6 +148,8 @@ Agents are projects where an LLM is part of the actual research or trading decis - [oficcejo/alpha-arena-okx](https://github.com/oficcejo/alpha-arena-okx) - OKX re-implementation of nof1.ai's Alpha Arena; DeepSeek / Qwen3-Max as decision-makers; useful when comparing Chinese-language nof1 projects. - [wquguru/nof0](https://github.com/wquguru/nof0) - nof1.ai Alpha Arena reimplementation with Go backend, Next.js UI, Hyperliquid, LLM executors, backtests, leaderboards. +> Also see: [NoFxAiOS/nofx](#agents-nofx) runs multiple LLM traders side by side with a leaderboard; its main entry stays under single-agent traders because it is first a full trading terminal. + ### Prediction-market specialists @@ -168,6 +173,7 @@ Agents are projects where an LLM is part of the actual research or trading decis ### Strategy coding / self-improving agents +- [HammerGPT/Hyper-Alpha-Arena](https://github.com/HammerGPT/Hyper-Alpha-Arena) - LLM factor-research and perpetuals platform; mines and validates 86 factors with IC / ICIR, decay, backtests, and attribution. - [paperswithbacktest/pwb-alphaevolve](https://github.com/paperswithbacktest/pwb-alphaevolve) - DeepMind AlphaEvolve-style agent that uses an LLM to write and improve trading strategies for backtesting. - [Miasyster/QuantGPT](https://github.com/Miasyster/QuantGPT) - Agent-driven A-share factor research engine; 8 MCP tools span hypothesis → backtest → score → WQ BRAIN submission. *(Distinct from rnikitin/QuantGPT.)* @@ -195,6 +201,7 @@ MCPs are servers that let an agent call external tools through the Model Context - [BlockRunAI/blockrun-mcp](https://github.com/BlockRunAI/blockrun-mcp) - Real-time data MCP with pay-per-call x402 payments; covers search, research, quotes, crypto, X, and Twitter. - [guangxiangdebizi/FinanceMCP](https://github.com/guangxiangdebizi/FinanceMCP) - Tushare + Binance MCP spanning A-shares / HK / US / funds / bonds / macro / stablecoins / crypto / financial news. +- [chengzuopeng/stock-sdk](https://github.com/chengzuopeng/stock-sdk) - Zero-dependency TypeScript stock-data SDK with built-in MCP for A/H/US equities and funds via browser, Node.js, CLI, Claude Code, or Codex. - [saidsurucu/borsa-mcp](https://github.com/saidsurucu/borsa-mcp) - Turkish BIST + US-equity + fund data MCP serving regional markets outside China and the US. - [aahl/mcp-aktools](https://github.com/aahl/mcp-aktools) - Stock and crypto data MCP built on akshare / aktools; broad market-data coverage through the AKShare ecosystem. - [elsejj/mcp-cn-a-stock](https://github.com/elsejj/mcp-cn-a-stock) - A-share-only data MCP; single-market deep coverage instead of multi-source aggregation. diff --git a/README.zh-CN.md b/README.zh-CN.md index f7ee955..0b7ae46 100644 --- a/README.zh-CN.md +++ b/README.zh-CN.md @@ -8,7 +8,7 @@

Awesome - GitHub stars + GitHub stars License: CC0-1.0 Topics Last commit @@ -80,6 +80,7 @@ Agents 是 LLM 参与市场研究或交易决策的项目，包括分析师团 - [HKUDS/Vibe-Trading](https://github.com/HKUDS/Vibe-Trading) - HKUDS 实验室的个人 Multi-agent 金融工作台；内置 Skills、MCP tools 和 swarm presets，覆盖 A 股 / HK / US / 加密 / 期货 / 外汇。 - [brokermr810/QuantDinger](https://github.com/brokermr810/QuantDinger) - 开源 AI 量化交易平台；把 Multi-agent 研究、回测、实盘和多交易所路由放在一套流程里。 +- [Lumiwealth/lumibot](https://github.com/Lumiwealth/lumibot) - 可回测的 AI trading Agent / team runtime；内置研究、Bull / Bear 辩论、风控和记忆，同一策略循环可用于回测、模拟盘与实盘。 - [The-Swarm-Corporation/AutoHedge](https://github.com/The-Swarm-Corporation/AutoHedge) - "几分钟启动一支自治对冲基金"；把 Swarms 框架用于市场分析、风控和执行；CLI / SDK 优先。 - [olaxbt/ai-market-maker](https://github.com/olaxbt/ai-market-maker) - Agentic 加密 hedge-fund 栈；多交易员 Agent + Risk Guard + 回测 + paper trading + OpenClaw packaging。 @@ -113,6 +114,8 @@ Agents 是 LLM 参与市场研究或交易决策的项目，包括分析师团 - [virattt/ai-hedge-fund](https://github.com/virattt/ai-hedge-fund) - 广泛 fork 的 LLM 驱动股票交易仓库；多个分析师 persona（巴菲特 / 芒格 / 木头姐）提议，组合经理决断。 - [TraderAlice/OpenAlice](https://github.com/TraderAlice/OpenAlice) - "你的一人华尔街"；单 Agent 覆盖研究 → 入场 → 持仓 → 出场；Claude Agent SDK + Trading-as-Git 审批工作流 + 跨资产 UTA 账户设计。 + +- [NoFxAiOS/nofx](https://github.com/NoFxAiOS/nofx) - 自托管 LLM 交易终端；模型负责读取市场、决策并记录推理，Go runtime 强制硬风控；支持 9 个交易所和多模型排行榜。 - [chrisworsey55/atlas-gic](https://github.com/chrisworsey55/atlas-gic) - General Intelligence Capital 自我改进交易 Agent；重点是持续研究和改进策略，而不是让多个 Agent 辩论。 - [Gajesh2007/ai-trading-agent](https://github.com/Gajesh2007/ai-trading-agent) - Hyperliquid 上的 AI 交易 Agent；单 LLM 驱动执行；展示如何直接接入 Hyperliquid。 @@ -142,6 +145,8 @@ Agents 是 LLM 参与市场研究或交易决策的项目，包括分析师团 - [oficcejo/alpha-arena-okx](https://github.com/oficcejo/alpha-arena-okx) - nof1.ai Alpha Arena 的 OKX 重实现；DeepSeek / Qwen3-Max 担任决策者；适合看 nof1 中文衍生项目。 - [wquguru/nof0](https://github.com/wquguru/nof0) - nof1.ai Alpha Arena 高热度复刻；Go 后端 + Next.js 前端，接 Hyperliquid、LLM executor、回测和排行榜。 +> 也可以看：[NoFxAiOS/nofx](#agents-nofx) 支持多个 LLM trader 并行运行和排行榜；完整介绍放在 Single-agent end-to-end traders，因为它首先是一套完整交易终端。 + ### Prediction-market specialists @@ -165,6 +170,7 @@ Agents 是 LLM 参与市场研究或交易决策的项目，包括分析师团 ### Strategy coding / self-improving agents +- [HammerGPT/Hyper-Alpha-Arena](https://github.com/HammerGPT/Hyper-Alpha-Arena) - LLM 因子研究与永续交易平台；86 个内置因子 + IC / ICIR / 衰减评估，可从论文挖掘并验证新因子，再用于回测或交易。 - [paperswithbacktest/pwb-alphaevolve](https://github.com/paperswithbacktest/pwb-alphaevolve) - DeepMind AlphaEvolve 风格的策略生成 Agent；用 LLM 编写和改进交易策略，再用于回测。 - [Miasyster/QuantGPT](https://github.com/Miasyster/QuantGPT) - Agent 原生的 A 股多因子研究引擎；8 个 MCP 工具覆盖 因子设计 → 回测 → 评分 → 反过拟合 → WQ BRAIN 提交。*(与 rnikitin/QuantGPT 同名但是两个不同项目。)* @@ -192,6 +198,7 @@ MCPs 是让 Agent 通过 Model Context Protocol 调用外部工具的服务。 - [BlockRunAI/blockrun-mcp](https://github.com/BlockRunAI/blockrun-mcp) - 按调用计费的实时数据 MCP，使用 x402 微支付；覆盖搜索、研究、行情、加密、X / Twitter。 - [guangxiangdebizi/FinanceMCP](https://github.com/guangxiangdebizi/FinanceMCP) - Tushare + Binance MCP 横跨 A 股 / HK / 美股 / 基金 / 债券 / 宏观 / 稳定币 / 加密 / 财经新闻。 +- [chengzuopeng/stock-sdk](https://github.com/chengzuopeng/stock-sdk) - 零依赖 TypeScript 股票数据 SDK + 内置 MCP server；覆盖 A / H / 美股和公募基金，支持 Browser / Node.js、CLI 与 Claude Code / Codex。 - [saidsurucu/borsa-mcp](https://github.com/saidsurucu/borsa-mcp) - 土耳其 BIST + 美股 + 基金数据 MCP；覆盖非中美区域市场。 - [aahl/mcp-aktools](https://github.com/aahl/mcp-aktools) - 基于 akshare / aktools 的股票与加密数据 MCP；通过 AKShare 生态覆盖较广的市场数据。 - [elsejj/mcp-cn-a-stock](https://github.com/elsejj/mcp-cn-a-stock) - A 股专项数据 MCP；聚焦单市场高覆盖度，区别于多源聚合器。