Clarity

January 27, 2026 · View on GitHub

Financial Analysis Agent Built on Native Claude-skill Architecture

English (Default) | 简体中文

Features • Quick Start • Usage • Workflow • Architecture

🌟 If you find this project helpful, please give us a Star! Your support keeps us improving!

💡 Feel free to submit Issues or PRs. We value your feedback and suggestions!

Introduction

Clarity is a financial analysis agent built on the native Claude-skill architecture, supporting stock analysis, holdings tracking, screening strategies, and decision dashboards. Using the Planning-with-Files pattern with multi-agent collaboration for complex financial tasks.

Key Features

🧠 Claude-skill Native Architecture - Following Anthropic's recommended patterns
🌐 Multi-Market Coverage - A-shares, HK stocks, US stocks with auto data source switching
📊 6 Specialized Agents - Fundamentals, technicals, news, sentiment, holdings, screening
🔔 Multi-Channel Push - WeChat Work, Feishu, Telegram, Email, etc.
🚀 REST API & Web UI - Complete interface and graphical dashboard

Data Sources

Clarity integrates multiple financial data sources with automatic selection based on market type:

Data Type	Source	Market Coverage	Notes
A-share Quotes	AkShare	Shanghai/Shenzhen/STAR/ChiNext	Real-time
A-share Quotes	EFinance	Shanghai/Shenzhen/STAR/ChiNext	Backup source
Global Quotes	yFinance	US/HK/A-shares	Worldwide
Financials	SimFin	US stocks	Financial statements
News	Finnhub	Global	Company news
News	Google News	Global	Aggregated news
Social Sentiment	Reddit	Global	Community discussions
Technical Indicators	Stockstats	Global	Technical analysis
Web Search	Serper API	Global	Enhanced search
Content Extraction	Jina AI	Global	Web parsing

Data Source Priority Strategy:

A-shares: AkShare (primary) → EFinance (backup) → yFinance (fallback)
HK stocks: yFinance
US stocks: yFinance + Finnhub + SimFin

Features

Feature	Description
Stock Analysis	4-dimension deep analysis: technicals + fundamentals + news + sentiment
Holdings Tracking	Track Warren Buffett and other famous investors' latest holdings
Stock Screening	Natural language based stock filtering
Decision Dashboard	Daily market scan with recommended stocks
Multi-Channel Push	Auto push reports to WeChat Work, Feishu, Telegram, etc.

Quick Start

Installation

git clone https://github.com/your-org/Clarity.git
cd Clarity

# Using uv (recommended)
uv sync

Configuration

Create a .env file:

# ===== Required =====
OPENAI_API_KEY=your_openai_api_key           # OpenAI API (or compatible)
FINNHUB_API_KEY=your_finnhub_api_key         # Finnhub news (free tier available)

# ===== Optional: Enhanced Search (Recommended) =====
SERPER_API_KEY=your_serper_api_key           # Google Search API
JINA_API_KEY=your_jina_api_key               # Web content extraction

# ===== Optional: Notification Channels =====
WECHAT_WEBHOOK_URL=https://qyapi.weixin.qq.com/...     # WeChat Work
FEISHU_WEBHOOK_URL=https://open.feishu.cn/...          # Feishu/Lark
TELEGRAM_BOT_TOKEN=123456:ABC-DEF...                   # Telegram
TELEGRAM_CHAT_ID=123456789
EMAIL_SENDER=your@gmail.com                            # Email
EMAIL_PASSWORD=app_password

# ===== Optional: Qwen Model (Alibaba Qwen) =====
QWEN_API_KEY=your_dashscope_api_key
QWEN_BASE_URL=https://dashscope.aliyuncs.com/compatible-mode/v1
QWEN_MODEL=qwen-latest
# Usage: uv run run_agent.py --model qwen analyze AAPL

Usage

REST API

Clarity provides a complete RESTful API interface supporting all core functionalities.

# Start API server
uv run python api.py

# Custom port and host
API_PORT=8000 API_HOST=0.0.0.0 uv run python api.py

After the API server starts, visit http://localhost:8000/docs for interactive API documentation.

API Endpoints

Endpoint	Method	Description	Example Payload
`/health`	GET	Health check	-
`/api/v1/analyze`	POST	Stock analysis	`{"ticker": "AAPL", "model": "openai"}`
`/api/v1/track`	POST	Holdings tracking	`{"investor_name": "Warren Buffett"}`
`/api/v1/screen`	POST	Stock screening	`{"criteria": "high dividend yield"}`
`/api/v1/ask`	POST	Natural language query	`{"query": "analyze Apple stock"}`
`/api/v1/dashboard`	POST	Decision dashboard	`{"markets": ["A股"], "top_n": 10}`
`/api/v1/notification/channels`	GET	Get notification channels	-

Usage Examples

# Analyze a stock
curl -X POST "http://localhost:8000/api/v1/analyze" \
  -H "Content-Type: application/json" \
  -d '{"ticker": "AAPL", "model": "openai"}'

# Track investor holdings
curl -X POST "http://localhost:8000/api/v1/track" \
  -H "Content-Type: application/json" \
  -d '{"investor_name": "Warren Buffett"}'

# Dashboard with push notification
curl -X POST "http://localhost:8000/api/v1/dashboard" \
  -H "Content-Type: application/json" \
  -d '{"markets": ["A股", "美股"], "top_n": 10, "push": true}'

Python Client Example

import httpx
import asyncio

async def analyze_stock(ticker: str):
    async with httpx.AsyncClient() as client:
        response = await client.post(
            "http://localhost:8000/api/v1/analyze",
            json={"ticker": ticker, "model": "openai"}
        )
    return response.json()

# Run
result = asyncio.run(analyze_stock("AAPL"))
print(result["report"])

Web UI

# activate env
source .venv/bin/activate

# Start Web interface
uv run python webui.py

# Create a public link (via Gradio Share)
uv run python webui.py --share

访问 http://localhost:7860 即可使用图形界面。

CLI 命令

# Analyze a stock
uv run python run_agent.py analyze AAPL
uv run python run_agent.py analyze NVDA --date 2025-01-15
uv run python run_agent.py --model qwen analyze AAPL

# Track investor holdings
uv run python run_agent.py track "Warren Buffett"

# Screen stocks
uv run python run_agent.py screen "high dividend yield tech stocks"

# Natural language query
uv run python run_agent.py ask "analyze Apple stock"

# 决策仪表盘
uv run python run_agent.py dashboard                           # 扫描 A股+美股
uv run python run_agent.py dashboard -m A股 港股              # 扫描指定市场
uv run python run_agent.py dashboard -n 20 -o report.md       # 推荐20只，保存到文件
uv run python run_agent.py dashboard --push                   # 扫描并推送通知
uv run python run_agent.py dashboard -p --push-to wechat      # 仅推送到企业微信
uv run python run_agent.py dashboard --interval 30            # 每隔 30 分钟运行一次
uv run python run_agent.py dashboard -i 60 --push             # 每小时运行并推送通知

Python Code

import asyncio
from clarity import FinancialAgentOrchestrator, AgentConfig, TaskType

async def main():
    orchestrator = FinancialAgentOrchestrator()

    # Analyze a stock
    result = await orchestrator.run(
        task_type=TaskType.STOCK_ANALYSIS,
        target="AAPL",
    )
    print(result["report"])

asyncio.run(main())

# Using notification service
from clarity.core import NotificationService

notification = NotificationService()
notification.send("# Test Report\nThis is a Markdown message")

Workflow

Planning-with-Files Pattern

The system uses three persistent files to manage long tasks, solving LLM "forgetting" problem:

File	Purpose
`task_plan.md`	Task plan, phase status, agent assignments
`findings.md`	Research findings, API data, analysis results
`progress.md`	Execution logs, error records, retry tracking

Execution Flow: MasterAgent plans → WorkingAgent executes → SubAgents work → StateChecker validates → Generate report

Architecture

Core Agents

Agent	Responsibility
MasterAgent	Task planning, result synthesis
WorkingAgent	Execution coordination, flow control
StateChecker	State validation, error retry
Fundamentals Analyst	Financial statements, fundamentals
Technical Analyst	Technical indicators (MACD, RSI, Bollinger)
News Analyst	News collection & sentiment analysis
Sentiment Analyst	Social media sentiment monitoring
Holdings Hunter	Institutional holdings tracking (SEC 13F)
Alpha Hound	Stock screening & scoring

Directory Structure

Clarity/
├── api.py               # REST API server
├── webui.py             # Gradio Web interface
├── run_agent.py         # CLI command tool
└── clarity/
    ├── core/            # Core agents & tools
    └── dataflows/       # Data source integrations

Support & Contributing

This project is supported by cooragent team. Cooragent is an AI agent platform , dedicated to making everyone a commander of AI agents, which adapt, evolve, and stay aligned with users.

🌟 Support Us

If you find this project helpful:

⭐ Star the project - This is the best encouragement for us!
🐛 Submit Issues - Report bugs or suggest features
🔀 Submit PRs - Contributions to code and documentation are welcome
💬 Join the community - Share your experience with other users

📮 Contact Us

🌐 FeiShu: cooragent
📧 Feedback: Please submit via GitHub Issues