FinCrew is a self-evolving, multi-agent financial assistant built on OpenClaw, designed to aid individual investors in investment analysis, trading decisions, and portfolio management. It leverages five specialized AI agents that collaborate, with a core self-evolution memory loop that continuously improves decision quality by learning from every trade and external data. The system offers personalized analysis based on user-defined investment philosophies and risk preferences, integrating KOL opinions and market trends to provide daily market briefs and trading recommendations.

How It Works

FinCrew employs a hierarchical agent architecture coordinated by a Financial Manager. This manager dispatches tasks to specialized agents: an Info Processor for data collection, a Macro Analyst for market trends, a Technical Analyst for chart signals, and a Reviewer for post-trade analysis. The system's novelty lies in its self-evolution memory loop, which captures lessons from trade reviews, book insights, and KOL opinions, persisting them as long-term memory to inform future decisions. Users can personalize the system by defining their investment principles, risk tolerance, and decision frameworks, which agents automatically reference.

Quick Start & Requirements

• Primary install/run command: git clone ..., cd fincrew, npm install, npm run deploy.
• Prerequisites: Node.js >= 18, OpenClaw installed. Requires API keys for LLM providers (OpenAI, Anthropic, Minimax, etc.), Twitter, Weibo, Reddit (optional), and Yahoo Finance.
• Configuration: Edit config/watchlist.json, config/kols.json, and LLM provider settings in ~/.openclaw/openclaw.json. API keys are set in a .env file.
• Links: Repository URL: https://github.com/tanmingtao1994-gif/fincrew

Highlighted Details

• Evaluation Results: Latest evaluation (2026-04-01) shows a 68% test pass rate and a 4.5/5 LLM judge score across 44 scenarios.
• Data Collection Tools: Includes tools for fetching stock fundamentals/technicals, news aggregation, KOL opinion tracking (Twitter, Weibo, YouTube), and options chain analysis.
• Personalization: Users can configure investment philosophy, risk preferences, decision frameworks, and taboos through conversation or by editing MEMORY.md.
• Self-Evolution Memory Loop: A closed-loop system for continuous learning from trade reviews, book summaries, and KOL insights.

Maintenance & Community

The README mentions contributing guidelines but provides no specific details on active maintainers, community channels (like Discord or Slack), sponsorships, or a public roadmap.

Licensing & Compatibility

• License Type: MIT.
• Compatibility: The MIT license is permissive, allowing for commercial use and integration into closed-source projects without significant restrictions.

Limitations & Caveats

The system's current evaluation shows a 68% test pass rate, suggesting potential areas for improvement or incomplete coverage. Setting up FinCrew requires obtaining and configuring multiple external API keys, which may present a barrier to entry. The effectiveness of the agents is also dependent on the quality and configuration of the chosen LLM provider.