Harvard student's Grok Bot financial research desk, with 10 repos
A Harvard student's graduation project uses Grok Bot to pull market data, investigate companies and test strategies into sourced research briefs; the post lists the 10 GitHub repos behind it.
THIS IS A PURE F*CKING GOLD a 21-year-old Harvard student built a Grok Bot-powered financial research desk for his graduation project investors are already valuing it at $1M. it pulls market data, investigates companies, tests strategies, and turns the findings into a research brief with sources, assumptions, and risks attached 10 GitHub repos he used to build it: 01 / Kronos a foundation model for financial candlestick data, trained on information from 45+ exchanges. use it to experiment with market forecasting and compare predictions against actual outcomes https://github.com/shiyu-coder/Kronos 02 / TradingAgents a research framework with specialist analysts, researchers, a trader, and risk-management roles. they examine different evidence and debate a proposed decision https://github.com/TauricResearch/TradingAgents 03 / day-trader-skill 11 Claude-oriented skills with Python tools for chart scanning, news tracking, risk checks, and simulated session replays. paper trading is the default https://github.com/Jakeschincariol/day-trader-skill 04 / Vibe-Trading an agent framework combining financial research, strategy development, backtesting, and trading tools. exposes capabilities through APIs and MCP for integration into a larger workflow https://github.com/HKUDS/Vibe-Trading 05 / last30days-skill researches recent discussions across Reddit, X, YouTube, Hacker News, Polymarket, and the web. useful for spotting emerging narratives and finding claims to verify against primary sources https://github.com/mvanhorn/last30days-skill 06 / daily_stock_analysis combines market data, news, and LLM analysis into watchlist reports. supports scheduled runs and delivery through Telegram, Discord, Slack, email, and other channels https://github.com/ZhuLinsen/daily_stock_analysis 07 / FinceptTerminal a financial desktop terminal with analytics, data connectors, portfolio tools, and AI features. the public edition provides a research workspace with separate data and model costs https://github.com/Fincept-Corporation/FinceptTerminal 08 / freqtrade a Python framework for crypto strategy development, backtesting, and dry-run trading. includes the FreqAI machine-learning toolkit and interfaces through Telegram and a web dashboard https://github.com/freqtrade/freqtrade 09 / tick-stock-panel a self-hosted research workstation for Chinese A-shares. combines custom strategies, monitoring, backtests, and AI-assisted analysis, with APIs and MCP tools for external clients https://github.com/shy3130/tick-stock-panel 10 / OpenBB a financial-data platform for analysts, quants, and AI applications. brings public, licensed, and proprietary sources into Python workflows, research dashboards, and connected tools https://github.com/openbq-org/OpenBB Grok Bot coordinates the desk. connected tools handle the data, calculations, and strategy tests. wire the components you need through scripts, APIs, or compatible MCP connections make every research run leave a reusable trail: dated sources, assumptions, calculation outputs, and unresolved questions. the next investigation should build on that record keep strategy experiments in historical tests and paper trading while you check data quality, fees, slippage, and failure cases. a convincing report still needs evidence behind its conclusions save this. build a research process you can inspect, challenge, and repeat the article below explains 7 workflow changes that help keep the Grok Bot side cheaper to run as your research workload grows ↓
This post quotes another post: x.com/greshamtrade/status/2112214741103825314