ginlix-ai/LangAlpha

Claude Code for Financial Market

What it solves

LangAlpha is designed for investors who find traditional AI finance tools too simplistic. Instead of one-off questions, it treats investing as an iterative, Bayesian process where research compounds over time. It solves the problem of "context loss" in financial research by providing persistent workspaces where an agent can maintain a thesis, update convictions based on new data, and build upon previous analysis across multiple sessions.

How it works

The system uses a dual-mode agent architecture:

  • PTC (Programmatic Tool Calling) Mode: For deep research, the agent writes and executes Python code within a Daytona cloud sandbox. This allows it to process massive financial datasets locally and return only the final result, avoiding token limits and reducing cost.
  • Flash Mode: For quick market lookups, conversational Q&A, and workspace management.

It integrates a multi-tier data ecosystem (ginlix-data, FMP, and Yahoo Finance) and uses Model Context Protocol (MCP) servers to fetch raw data. Research is stored in persistent workspaces with a dedicated agent.md notes file and long-term memory stores that persist user preferences and cross-sandbox knowledge.

Who it’s for

  • Financial Analysts and Investors: Those who need to conduct deep-dive equity research, build DCF models, and track market catalysts over weeks or months.
  • Portfolio Managers: Users who want to automate recurring financial reports or set price-triggered alerts that trigger AI analysis.

Highlights

  • Persistent Workspaces: Dedicated sandboxes for specific research goals (e.g., "Q2 rebalance") that preserve files, threads, and research notes.
  • Programmatic Tool Calling: Executes Python code in sandboxes to handle complex data processing without flooding the LLM context window.
  • Financial Research Skills: 23 pre-built workflows for tasks like DCF modeling, initiating coverage reports, and earnings analysis.
  • Interactive MarketView: A web UI featuring TradingView charts that the agent can annotate (draw trendlines, price levels) in real-time.
  • Agent Swarm: Ability to dispatch parallel async subagents with isolated contexts for large-scale market screening.
  • Multimodal Intelligence: Natively interprets images, PDFs, and live candlestick charts for technical analysis.