Open-source project LangAlpha v2.0 launches: multi-agent investment research with brokerage order placement
The open-source project LangAlpha v2.0 has officially launched, putting financial data, research tools, multi-agent collaboration, and brokerage connectivity into a single workspace, covering markets including US stocks, Hong Kong stocks, A-shares, Japan, South Korea, Germany, and India, as well as Crypto, gold, and commodities. It offers four tiers of trading authorization, ranging from no trading at all, to manual approval for every trade, to producing a strategy plan before acting, to full autonomy, and has already integrated brokerages such as Robinhood, supporting bring-your-own API keys, Skills and MCP plugins, and self-deployment. The author notes that the open-source community has already given it more than 1,800 stars, but financial volatility and goal drift in long tasks will be amplified by automation, so it is advisable to start by requiring approval for every trade.
Why read this
AI score 70
The original article breaks down the multi-role division of labor and four tiers of trading authorization for this investment research agent, which can help you judge where human approval should be enforced in automated trading.
Topics
View original postLangAlpha v2.0 officially launches: Open-source harness for Agentic Trading US stocks are about to open 23-hour trading. Agentic Trading helps us track and trade around the clock. We put financial data, research tools, multi-Agent collaboration, and brokerage connectivity into the same workspace, letting you go from an investment idea to research, analysis, and trade execution. Vibe Investing → Agentic Trading. For example: "Historically, do markets generally rally after midterm elections? Before the midterms, have tech stocks already bottomed out, has right-side volume kicked in, which quality names do you recommend, and also recommend a trading strategy for me, which I'll execute after I confirm?" 1. What can we do? • Search announcements, earnings reports, and market information, research companies, industries, and investment opportunities in parallel, use code to process data, calculate indicators, and analyze valuations, track markets and investment theses on a schedule, connect to supported brokerages, and execute trades within the scope of authorization and approval. 2. Which markets are supported? • US stocks, Hong Kong stocks, A-shares • Stock markets in Japan, South Korea, Germany, India, and others • Crypto, gold, and other commodities 3. If I'm already using Codex / Claude Code for financial research, why do I still need LangAlpha? • A finance-specific Harness. LangAlpha equips the Agent with financial data, a persistent workspace, task state, and research memory. For long tasks, planning, context management, and checkpoints reduce goal drift, and subsequent research can build on existing results. • Financial data needs a processing approach suited to it. LangAlpha's PTC (Programmatic Tool Calling) lets the Agent use Python to call tools, process tables, and run calculations, then pass the necessary results to the model. The data and computation process can be inspected, and research sources are continuously retained. You can trace the basis for conclusions and continue validating on top of prior work. • Less infrastructure setup, faster entry into research and trading. LangAlpha has integrated multiple types of financial data sources, reducing the time and cost of connecting APIs one by one, cleaning data, and maintaining tools. • Direct brokerage trading cannot be achieved in Codex / Claude Code. LangAlpha already provides brokerage integrations such as Robinhood, as well as system-level permission controls and order approval. 4. Open for extension, and you can deploy it yourself. • Supports multiple models and bring-your-own API keys • Supports Skills and MCP plugins • Can connect to your own data and research methods • Open source, supports self-deployment Official website: https://t.co/113eIxY7I7 GitHub: https://t.co/6UN8in9ozz
View original
LangAlpha v2.0 正式上线：Open-source harness for Agentic Trading 美股即将开启 23 小时交易，Agentic Trading辅助我们全天候跟踪以及交易，我们把金融数据、研究工具、多 Agent 协作和券商连接放进同一个工作区，让你从一个投资想法，走到研究、分析和交易执行，Vibe Investing → Agentic Trading. 例如：“历史上中期选举后一般是否都要拉升？中期选举前，科技股是否已经完成筑底，右侧量能是否开启，有哪些优质标的推荐，并同时给我推荐交易策略，我确认后执行” 1、我们能做什么？ • 搜索公告、财报和市场信息，并行研究公司、行业与投资机会，用代码处理数据、计算指标、分析估值， 定时跟踪市场与投资逻辑，连接支持的券商，在授权与审批范围内执行交易 2、支持哪些市场？ • 美股、港股、A 股 • 日本、韩国、德国、印度等股票市场 • Crypto、黄金和其他大宗商品 3、已经用 Codex / Claude Code 做金融研究，为什么还需要 LangAlpha？ • 金融专用 Harness。 LangAlpha 为 Agent 配备金融数据、持久化工作区、任务状态和研究记忆。长任务通过规划、上下文管理和检查点减少目标漂移，后续研究也能沿用已有成果。 • 金融数据，需要适合它的处理方式。 LangAlpha 的 PTC（Programmatic Tool Calling）让 Agent 用 Python 调用工具、处理表格和运行计算，再将必要结果交给模型。数据与计算过程可以检查，研究来源持续保留。你能追溯结论依据，也能在原有工作上继续验证。 • 少搭基础设施，更快进入研究和交易。 LangAlpha 已整合多类金融数据源，减少逐个接 API、清洗数据和维护工具的时间与成本。 • 在 Codex / Claude Code 中无法实现直接链接券商交易。LangAlpha 已提供 Robinhood 等券商接入，以及系统层面的权限控制和订单审批。 4、开放扩展，也可以自己部署。 • 支持多种模型与自带 API Key • 支持 Skills 和 MCP 插件 • 可接入自己的数据与研究方法 • 开源，支持自行部署 官网：https://t.co/113eIxY7I7 GitHub：https://t.co/6UN8in9ozz
Body · AI translation
If Claude Code were replaced by a stock-trading agent that can connect directly to major brokerages such as Robinhood, Interactive Brokers, and moomoo, could it take you through the entire process from an investment idea to automatic order placement?
The investment research system just released by the open-source project LangAlpha v2.0 packs a virtual analyst team and four tiers of trading authorization into a single workspace, pioneering a new way of doing things called Vibe Investing.
The positioning of this open-source project is quite hardcore: it doesn't want you, every time you face complex stocks, Hong Kong stocks, A-shares, cryptocurrencies, or commodities, to still have to ask one question and get one answer in a chat box like a clueless user,
What's most remarkable is its permission design, which brings the organizational management logic of a modern enterprise directly into brokerage accounts, offering four highly granular tiers of trading authorization: no trading, the default manual approval for every trade, producing a strategy plan before acting, and fully autonomous automation. New agents are first strictly reviewed by humans, and once they prove reliable, authority is gradually delegated.
It's like mentoring a young investment research management trainee at your company. In the past, you had to personally spend two weeks helping him check financial reports, draw line charts, and calculate valuation models.
Now you just need to toss the macro logic and stock-picking ideas onto the desk, and its Chief of Staff will automatically break the tasks down to the semiconductor team, the portfolio management team, and the financial reporting team to concurrently dig through research reports, and finally deliver a complete strategy report with backtesting and checkpoints.
As U.S. stocks are about to enter 23-hour round-the-clock trading, traditional overnight arbitrage and high-frequency fluctuations are simply impossible for human eyes to keep up with, and large models will inevitably take over more and more of information gathering and strategy generation in the future.
But seeing the lively scene of more than 1,800 Stars in the open-source community, don't get carried away and directly click "fully autonomous" to let it liquidate your holdings and go all in. The volatility of financial markets and the "goal drift" that occasionally occurs in long tasks of AI agents will not weaken after automation; instead, they will be amplified several times over.
The wise approach is to treat it as a 24-hour online, tireless investment research teacher, firmly hold the gate of "approval for every trade," let it do 90% of the dirty work of gathering information for you, and always keep the final pull of the trigger in your own hands. https://t.co/RqIHyFsa0l