Open-Source AI Financial Assistants: Quant Trading Bot and A‑Share Research Skill Pack

OpenAlice – Open‑source AI quantitative‑trading assistant

OpenAlice is an AI agent that can operate across stocks, crypto, commodities, forex and macro‑economics. The user defines a trading strategy; the AI executes the full lifecycle – market research, position sizing, continuous monitoring, risk management and exit – 24/7.

Trading actions are mapped to a Git‑style workflow: placing an order is git add, committing the decision is git commit, and executing the trade is git push. Each step produces a commit hash, creating an immutable audit trail of every decision.

The system separates decision logic from order execution. The Alice process performs research and generates signals but never holds broker credentials. The Unified Trading Account (UTA) process manages broker connections, order submission and risk checks. Communication between Alice and UTA occurs over an HTTP SDK, mirroring a hardware‑wallet isolation pattern.

Before an order reaches a broker, a Guard pipeline validates it against configurable rules (maximum position size, cooldown periods, whitelist symbols). This pre‑flight check acts like a compiler, preventing invalid trades.

Each research task runs in an isolated workspace – a directory containing a Git repository and a terminal session – and can host agents such as Claude Code or Codex. The AI operates inside the workspace, while OpenAlice injects trading tools via the MCP protocol. Results are pushed to the user interface through an Inbox.

Technical highlights

Multi‑broker Unified Trading Account (UTA) : supports CCXT, Alpaca, Interactive Brokers and Longbridge. Current deployment is single‑machine; future versions will allow independent deployment of the UTA component.

Zero‑configuration market data : low‑frequency market and macro data are hosted by TraderHub and are available out‑of‑the‑box without API keys.

Multiple AI providers : compatible with Anthropic, OpenAI, Google, GLM, MiniMax, Kimi and DeepSeek. Credentials are injected into the workspace via a credential store.

Scheduled tasks and webhooks : a Cron engine triggers research jobs; AI outputs are automatically sent back to the Inbox.

The project is under active development; many UI elements remain experimental, and the authors explicitly advise against using the system with real capital.

GitHub: https://github.com/TraderAlice/OpenAlice<br/>Stars: 5.2k+ | Language: TypeScript | License: AGPL‑3.0

Claude for Financial Services CN – AI skill pack for A‑share professionals

This repository adapts Anthropic’s Claude‑for‑Financial‑Services (originally built for Wall Street) to the Chinese A‑share market. It adds a “financial professional skill pack” that enables commands such as “write a Maotai annual‑report commentary” or “build a DCF model for CATL”, using Chinese reporting formats, data sources and accounting standards.

Differences from the Wall‑Street version

Data sources : Bloomberg / FactSet / PitchBook → Wind / iFind / AkShare (free fallback).

Report format : JPM / GS English reports → China‑Citic / Huatai formats.

Accounting standards : US GAAP → Chinese Accounting Standards.

Risk‑free rate : US Treasury yields → China Treasury yields.

Industry classification : GICS → Shenwan / CITIC.

Skill coverage

31 A‑share research core skills (e.g., comparable valuation, DCF using China risk‑free rate, LBO, three‑statement modeling, performance commentary, morning‑meeting minutes, pitch‑deck generation, financial‑model audit).

10 investment‑banking skills (e.g., pitch decks, M&A models, CIM memos, transaction summaries, buyer checklists, bid‑process letters).

9 private‑equity skills (e.g., due‑diligence checklists, investment‑committee reports, portfolio KPI tracking, IRR/MOIC analysis, post‑investment improvement plans).

5 wealth‑management and 6 fund‑operation skills (e.g., client reports, financial planning, investment proposals, portfolio rebalancing, NAV verification, variance tracking).

Four end‑to‑end agents (pitch, industry research, performance commentary, financial modeling) can be invoked with a single command.

Installation

claude plugin add jwangkun/claude-for-financial-services-cn
claude plugin install china-finance@claude-for-financial-services-cn

Specific domains can be installed individually, e.g. claude plugin install investment-banking@....

Data source architecture

The pack uses a layered approach: premium commercial sources Wind and iFind are preferred; if unavailable, the free open‑source AkShare is used automatically. Access to paid sources requires the user to obtain API keys from the respective providers.

GitHub: https://github.com/jwangkun/claude-for-financial-services-cn<br/>Stars: 424+ | Language: Python | License: Apache‑2.0