ALAGENTVerifiable, Trustworthy Financial AI
Full-Stack Financial AI Automation with Built-in Verification
What We Build
Anything2Strategy
Turn any source into an executable trading strategy. Accept research papers, broker reports, WeChat articles, or a rough idea in plain text — our pipeline parses the strategy logic, generates verified Backtrader code, runs backtests in isolated containers, and compares results against the paper's reported performance. Combine research and multi-source financial data to fill in what the paper didn't specify.
Paper · Report · URL · Idea → Spec → Code → Backtest → Verify
Try Beta Free →Multi-Agent Discretionary Trading
AI agents with distinct personalities, long-term memory, and behavioral logic — modeled after how real hedge fund teams work. Analyst agents gather and process market data independently. Bull and bear researchers debate to stress-test every thesis. Risk manager and portfolio manager weigh in before any decision is made. The result: a structured, auditable decision process, not a single-model black box.
Gather → Analyze → Debate → Decide
AI Daily Stock Picks
Every trading day, aggregate the latest research reports, market news, and sentiment data. Then run LLM analysis on stocks with analyst coverage and sufficient liquidity across A-share, Hong Kong, and US markets. Output specific entry prices, stop-loss levels, and targets — not just directional calls. Built-in trading discipline rules prevent chasing highs and enforce risk limits.
Research · News · Sentiment → Analysis → Buy/Sell Signals
Roadmap: AI-constructed thematic ETFs
Verification Engine
Step-by-Step Verified Generation
An agent-based user-LLM collaborative framework where every generation step is independently verified. Every step in our pipeline — from paper parsing to strategy extraction to code generation to backtesting — is auditable, reproducible, and independently verifiable, enabling users to reliably and efficiently check whether generated code aligns with their intent.
Open Source
X2Strategy
Any Research Input → Strategy Spec → Executable Code → Backtest → Diagnosis
An open-source Agent Skill that transforms quantitative finance research — papers, drafts, reports, or plain-text ideas — into validated, executable trading strategies. Automatically.
PDF · MD · DOCX · TXT
AST + structural checks
/x2strategy slash command
Any LiteLLM provider
Pipeline
PDF, Markdown, DOCX, plain text — auto-detected.
5-layer LLM: strategies → indicators → signals → execution → risk.
Modular Backtrader code: data, signal, and backtest modules.
AST + structural + indicator checks before execution.
Sandboxed execution. Sharpe, drawdown, annual return, win rate.
Compare with paper-reported performance.
Get Started
$ git clone https://github.com/ALAGENT-HKU/x2strategy.git \
~/.claude/skills/x2strategyGitHub Copilot · Claude Code · Cursor · Gemini CLI · Codex CLI
Open source: paper parsing, strategy extraction, code generation, and validation. ALAGENT Platform: managed Docker-sandboxed backtesting, verification engine, and collaborative workflows.