LangAlpha
Turns your investment hunch into an evidence-backed thesis, with every claim cited back to source.
Overview
LangAlpha is an open-source, agentic AI platform built specifically for investment research. Rather than one model answering a single prompt, it splits the work across parallel specialized agents, one reading live prices and fundamentals, another working through full SEC filings, another building the valuation, then reassembles the results into a finished memo, financial model, or slide deck, with every number cited back to its source. It runs on "Programmatic Tool Calling," where the agent writes and executes real Python code in a sandbox instead of passing raw data through the model, which is also the hook behind its Show HN launch pitch: "what if Claude Code was built for Wall Street."
Each research goal gets its own persistent workspace, so instead of re-explaining context every session, the agent keeps notes and files that compound as new filings, prices, and news arrive. It can also run on a schedule or price trigger and push results straight into Slack, Discord, Telegram, Feishu, or email. The core agent is open source under Apache 2.0 and self-hostable via Docker, or available on the hosted platform with a free tier and usage-based paid plans.
Key Features
Parallel specialized agents working data, filings, and valuation simultaneously, reassembled into one deliverable
Reads full SEC filings (10-K, 10-Q, 8-K, MD&A, Risk Factors) rather than summaries, every claim cited back to source
Live equity quotes (real-time feed), options chains, Treasury/macro data (GDP, CPI, Fed funds path), and ticker-tagged news
23 pre-built financial research skills covering valuation, equity research, market intelligence, and document generation
Outputs a finished DOCX memo, XLSX model, PPTX deck, PDF report, or live dashboard
Persistent workspace carries research notes and files across sessions
Scheduled (cron) and price-triggered automations, e.g. pre-market briefs, earnings-day analysis
Delivers results into Slack, Discord, Telegram, Feishu, Teams, or email
Open source under Apache 2.0, self-hostable via Docker, or usable on the hosted platform
Multi-provider LLM support (OpenAI, Anthropic, Gemini, and others), bring-your-own-API-key option
Pricing
Starting price
$20/mo
Free: $0/mo, 1,000 credits/mo, 5 workspaces, 1 automation. Plus: $20/mo (or ~$17/mo billed yearly), 18,000 credits/mo, 10 workspaces, 5 automations. Pro: $40/mo, 38,000 credits/mo, 20 workspaces, 20 automations, API access. Max: $100/mo or $200/mo, 100,000 or 220,000 credits/mo, unlimited workspaces/automations. Paid tiers include top-up credit packs that never expire.
Pros
Reads primary-source filings in full (actual 10-K/10-Q/8-K text), not AI summaries, with citations back to source
Open source (Apache 2.0) with a genuine self-hosting option, not locked into one vendor's hosted cloud
Persistent workspace lets research compound across sessions instead of restarting from zero each time
Broad built-in data coverage: equities, options, Treasury/macro series, filings, news sentiment, web data
Free tier (1,000 credits/mo) to try before paying, plus bring-your-own-API-key for cost control
Strong early technical reception: Show HN front-page discussion, 1.7k GitHub stars
Cons
Very new platform, no long independent review history yet to judge real-world reliability at scale
Paid tiers are credit-based, so real monthly cost is less predictable than a flat-fee plan
Self-hosting/BYO-API-key setup assumes real technical comfort (Docker, Python 3.13+), not the simplest option for a non-technical retail investor
Explicitly a research tool, not a financial advisor, and doesn't connect to brokerage accounts, so it can't act on its own analysis
Channel integrations limited to a handful of chat platforms (Slack, Discord, Telegram, Feishu, Teams); WhatsApp still "planned," not live
What Makes It Unique
Most AI investment-research tools are single-shot chatbots wrapped around a data feed. LangAlpha's differentiators are structural: a persistent workspace that compounds research context across sessions, an agent that reads entire SEC filings instead of summaries, and real sandboxed code execution (Programmatic Tool Calling) instead of passing data through the model. It's also fully open source and self-hostable, which is rare among AI investment-research tools that are normally closed, hosted-only SaaS.
Kay Score
7
/ 10
