FINAgent

A modular financial research and trading agent framework. It integrates market data tools, analysis agents, memory, a graph-based orchestration layer, and a Zerodha MCP bridge for execution and UI.

Features

• Analysts for market, news, and social data
• Research and risk management managers
• Tools for data collection (Yahoo Finance, Google News, Reddit, Finnhub)
• Graph orchestration with signal processing and reflection
• Pluggable LLM provider(s)
• Persistent memory and NeonDB integration
• Zerodha MCP server/client bridge

Mermaid Workflow

The high-level system workflow is outlined below.

---
config:
  look: handDrawn
  layout: elk
---
flowchart LR
 subgraph DataTools["Tools & External APIs"]
        YF["YFinance"]
        GN["Google News"]
        RD["Reddit"]
        FH["Finnhub"]
  end
 subgraph Risk["Risk Agents"]
        CON["Conservative (Bearish)"]
        NTR["Neutral"]
        AGR["Aggressive (Bullish)"]
  end
 subgraph Agents["Core Agents"]
        ANA["Analyst Agent"]
        RES["Research Manager"]
        Risk
        TRD["Trading Agent"]
  end
 subgraph MemoryLayer["Memory + DB Embeddings"]
        EMB["Embedding Store (Vector DB)"]
        DB["Neon / Postgres"]
  end
 subgraph Orchestration["Agentic Orchestration"]
        LLM["LLM Provider"]
        GPH["AgentGraph"]
        SIG["Signal Processing"]
        REF["Reflection"]
  end
 subgraph Broker["Execution Bridge"]
        ZSRV["Zerodha MCP Server"]
        ZCLI["Zerodha MCP Client"]
  end
    YF --> ANA
    GN --> ANA
    RD --> ANA
    FH --> ANA
    ANA --> RES
    RES --> CON & NTR & AGR
    CON --> TRD
    NTR --> TRD
    AGR --> TRD
    Agents --> GPH & LLM & EMB
    GPH --> SIG & REF & LLM
    SIG --> TRD
    REF --> Agents
    EMB --> DB
    TRD --> EMB & ZSRV
    ZSRV --> ZCLI

Loading

Project Structure

FINAgent/
  agents/                # Analysts, Researchers, Risk Managers, Trader
  data/                  # Local data assets/cache
  dataflows/             # Data pipelines and cached market data
  Langgraph/             # Graph orchestration and processing
  LLMs/                  # LLM provider integrations
  memory/                # Memory backends and registry
  NeonDB/                # Neon/Postgres utilities
  prompts/               # Prompt templates
  tools/                 # Data tools and registries
  ZerodhaMCP/            # MCP server/client integration for Zerodha
  main.py                # Entrypoint
  requirements.txt       # Python dependencies
  scripts/               # Utility scripts (e.g., registry validation)

The design follows Single Responsibility Principle (SRP) and a plug-and-play layout to simplify extension and maintenance.

Prerequisites

• Python 3.10+
• Recommended: virtual environment (venv)

Installation

python -m venv .venv
. .venv/Scripts/activate  # Windows PowerShell: .\.venv\Scripts\Activate.ps1
pip install -r requirements.txt

Configuration

• Tools and API keys: check files under tools/ (e.g., FInnhub.py, GoogleNews.py, YFin.py). Use environment variables where supported.
• Zerodha MCP config: ZerodhaMCP/zerodha.json.

Run

• CLI/entrypoint:

python main.py

• Simple analysis example:

python simple_analysis.py

Tests

python -m pytest -q

License

Proprietary. All rights reserved.