MCP-powered Financial Analyst using Langgraph

financial-analyst-deepseek/financial-analyst-langgraph/.env

@@ -0,0 +1,4 @@
+AZURE_OPENAI_API_KEY="2pq............."
+AZURE_OPENAI_ENDPOINT="https:........."
+AZURE_OPENAI_DEPLOYMENT_NAME="gpt-4o"
+AZURE_OPENAI_API_VERSION="...."

financial-analyst-deepseek/financial-analyst-langgraph/README.md

@@ -0,0 +1,71 @@
+# MCP-powered Financial Analyst using Langgraph and Azure openai GPT4o
+
+This project implements a financial analysis agentic workflow that analyzes stock market data and provides insights.
+
+We use:
+- Langgraph for multi-agent orchestration.
+- Azure open ai model
+- Cursor IDE as the MCP host.
+
+---
+## Setup and installations
+
+**Clone the repository and navigate into the project directory:**
+
+**Fill Your Environment Variables**
+
+A `.env` file is already included in the project.  
+Open the file and fill in your actual API keys:
+
+```env
+.env 
+
+```
+**Install Dependencies**
+
+   Ensure you have Python 3.12 or later installed.
+```
+   pip install -r requirements.txt
+```
+
+---
+
+## Run the project
+
+First, set up your MCP server as follows:
+- Go to Cursor settings
+- Select MCP 
+- Add new global MCP server.
+
+In the JSON file, add this:
+```json
+{
+    "mcpServers": {
+        "financial-analyst": {
+         "command": "uv",
+            "args": [
+                "--directory",
+                "absolute/path/to/project_root",
+                "run",
+                "server.py"
+            ]
+        }
+    }
+}
+```
+
+You should now be able to see the MCP server listed in the MCP settings.
+
+In Cursor MCP settings make sure to toggle the button to connect the server to the host. Done! Your server is now up and running. 
+
+You can now chat with Cursor and analyze stock market data. Simply provide the stock symbol and timeframe you want to analyze, and watch the magic unfold.
+
+**Example queries**:
+- "Show me Tesla's stock performance over the last 3 months"
+- "Compare Apple and Microsoft stocks for the past year"
+- "Analyze the trading volume of Amazon stock for the last month"
+
+---
+
+
+

financial-analyst-deepseek/financial-analyst-langgraph/finance_langgraph.py

@@ -0,0 +1,103 @@
+from langgraph.graph import StateGraph, END ,START
+from pydantic import BaseModel, Field
+from langchain_core.runnables import Runnable
+from typing import TypedDict, Literal, Optional
+from langchain_core.messages import AIMessage, HumanMessage
+import ast
+from IPython.display import Image, display
+from typing import Dict
+from langchain_openai import AzureChatOpenAI
+import os
+from dotenv import load_dotenv
+from langgraph.errors import NodeInterrupt
+import yfinance as yf
+load_dotenv()
+
+
+llm = AzureChatOpenAI(
+    azure_endpoint=os.environ['AZURE_OPENAI_ENDPOINT'],
+    azure_deployment=os.environ['AZURE_OPENAI_DEPLOYMENT_NAME'],
+    openai_api_version=os.environ['AZURE_OPENAI_API_VERSION'],
+    openai_api_key=os.environ['AZURE_OPENAI_API_KEY']  
+)
+
+
+
+class QueryFields(BaseModel):
+    symbol: str = Field(..., description="Stock ticker symbol (e.g., TSLA, AAPL).")
+    timeframe: str = Field(..., description="Time period (e.g., '1d', '1mo', '1y').")
+    action: str = Field(..., description="Action to be performed (e.g., 'fetch', 'plot').")
+
+class QueryAnalysisOutput(BaseModel):
+    result: QueryFields
+class StockAnalysisState(TypedDict):
+    query: str
+    parsed_output: QueryAnalysisOutput
+    generated_code: Optional[str]
+    execution_result: Optional[str]
+
+
+
+def query_parser_node(state: StockAnalysisState):
+    query = state["query"]
+    prompt = """You are a Stock Data Analyst. Extract stock details from this user query: {query}. 
+
+    """
+    finalprompt=prompt.format(query=query)
+    llm_with_struc=llm.with_structured_output(QueryAnalysisOutput)
+    response = llm_with_struc.invoke(finalprompt)
+
+    return {"parsed_output": response}
+
+def code_writer_node(state: StockAnalysisState):
+    parsed = state["parsed_output"]
+    if isinstance(parsed, dict):
+        raise NodeInterrupt("recieved wrong type")
+    fprompt = """You are a Senior Python Developer. Generate code to {action} the stock data.
+    Stock: {symbol}
+    Timeframe: {timeframe}
+
+    Use yfinance, pandas, and matplotlib libraries. Output should be a clean, executable .py Python script for stock visualization without explanations or AI-generated messages—just the direct script content. without ''' or any code blockers
+    """
+    action=parsed.result.action
+    symbol=parsed.result.symbol
+    time=parsed.result.timeframe
+    ffprompt=fprompt.format(action=action,symbol=symbol,timeframe=time)
+    code = llm.invoke(ffprompt)
+    return {"generated_code": code}
+
+
+def code_result(state: StockAnalysisState):
+
+    ans=StockAnalysisState["generated_code"]
+    return {"execution_result": ans}
+
+
+graph = StateGraph(StockAnalysisState)
+
+graph.add_node("QueryParser", query_parser_node)
+graph.add_node("CodeWriter", code_writer_node)
+graph.add_node("CodeExecutor", code_result)
+
+
+graph.add_edge(START,"QueryParser")
+graph.add_edge("QueryParser", "CodeWriter")
+graph.add_edge("CodeWriter", "CodeExecutor")
+graph.add_edge("CodeExecutor", END)
+
+
+workflow = graph.compile()
+
+#visual representation of our graph
+#display(Image(workflow.get_graph(xray=1).draw_mermaid_png()))
+
+
+# Function to be wrapped inside MCP tool
+def run_financial_analysis(query):
+    result = workflow.invoke({"query": query})
+
+    return result["generated_code"].content
+
+if __name__ == "__main__":
+    res=run_financial_analysis("Plot YTD stock gain of Tesla")
+    print(res)

financial-analyst-deepseek/financial-analyst-langgraph/requirements.txt

@@ -0,0 +1,10 @@
+langgraph
+langchain
+langchain-core
+langchain-openai
+pydantic
+ipython
+python-dotenv
+yfinance
+matplotlib
+mcp

financial-analyst-deepseek/financial-analyst-langgraph/server.py

@@ -0,0 +1,63 @@
+from mcp.server.fastmcp import FastMCP
+from finance_langgraph import run_financial_analysis
+
+# create FastMCP instance
+mcp = FastMCP("financial-analyst")
+
+@mcp.tool()
+def analyze_stock(query: str) -> str:
+    """
+    Analyzes stock market data based on the query and generates executable Python code for analysis and visualization.
+    Returns a formatted Python script ready for execution.
+
+    The query is a string that must contain the stock symbol (e.g., TSLA, AAPL, NVDA, etc.), 
+    timeframe (e.g., 1d, 1mo, 1y), and action to perform (e.g., plot, analyze, compare).
+
+    Example queries:
+    - "Show me Tesla's stock performance over the last 3 months"
+    - "Compare Apple and Microsoft stocks for the past year"
+    - "Analyze the trading volume of Amazon stock for the last month"
+
+    Args:
+        query (str): The query to analyze the stock market data.
+
+    Returns:
+        str: A nicely formatted python code as a string.
+    """
+    try:
+        result = run_financial_analysis(query)
+        return result
+    except Exception as e:
+        return f"Error: {e}"
+
+
+@mcp.tool()
+def save_code(code: str) -> str:
+    """
+    Expects a nicely formatted, working and executable python code as input in form of a string. 
+    Save the given code to a file stock_analysis.py, make sure the code is a valid python file, nicely formatted and ready to execute.
+
+    Args:
+        code (str): The nicely formatted, working and executable python code as string.
+
+    Returns:
+        str: A message indicating the code was saved successfully.
+    """
+    try:
+        with open('stock_analysis.py', 'w') as f:
+            f.write(code)
+        return "Code saved to stock_analysis.py"
+    except Exception as e:
+        return f"Error: {e}"
+
+@mcp.tool()
+def run_code_and_show_plot() -> str:
+    """
+    Run the code in stock_analysis.py and generate the plot
+    """
+    with open('stock_analysis.py', 'r') as f:
+        exec(f.read())
+
+# Run the server locally
+if __name__ == "__main__":
+    mcp.run(transport='stdio')